CN104615884B - A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method - Google Patents
A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method Download PDFInfo
- Publication number
- CN104615884B CN104615884B CN201510056900.2A CN201510056900A CN104615884B CN 104615884 B CN104615884 B CN 104615884B CN 201510056900 A CN201510056900 A CN 201510056900A CN 104615884 B CN104615884 B CN 104615884B
- Authority
- CN
- China
- Prior art keywords
- risk
- node
- infectious disease
- severe
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Medical Treatment And Welfare Office Work (AREA)
Abstract
Module, risk network struction module, risk assessment prediction module, grading forewarning system module are chosen the present invention relates to a kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method, including risk definition module, risk elements;The present invention can effectively solve tradition based on infectious disease spatial abnormal feature judge modes of warning judgement it is imprecise, not in time, be also easy to produce excessive early warning the problems such as;Supplemented and optimized available for the infectious disease early stage space-time modes of warning to the national monitoring of infectious disease automatic early-warning information platform of China.
Description
Technical field
, can root the present invention relates to a kind of infectious disease severe based on viral recall rate, mortality risk forecasting system and method
According to the recall rate of certain Causative virus of infectious disease, with reference to weather environment and risk population distribution situation, the infection is quickly determined
Sick following occur severe, dead risk, and thus targetedly sends early warning, available for supplementing and optimize state of China handed down from the older generations of the family
Catch an illness and monitor the infectious disease early stage space-time modes of warning of automatic early-warning information platform.
Background technology
In recent years, the environmental pollution being on the rise and scarcity of resources, the drug-resistant variants of original bio-pathogen, new hair are passed
The appearance caught an illness, the weak public health service ability in economically underdeveloped area and country, population of city districts it is excessive it is intensive,
The growth of increasingly frequent and some areas ultranationalism and the religion force that are exchanged between countries and regions so that the whole world
The period occurred frequently of a public health emergency is progressed into, the life security of the mankind receives serious threat.How to have
The harm that the strick precaution accident of effect is brought, it has also become the severe problem that national governments face.It is pre- that foundation includes infectious disease early stage
Public health emergency emergency reaction system including alert system be effective control outbreaks of infectious diseases with it is popular and effectively
Tackle the important measures of great public health event.The symptom of a trend of the health event that notes abnormalities only early is science, effective right
Early warning is made in " symptom of a trend " public health emergency, could be to implement various counter-measures to gain time, emerging public health
Event control will not cause crisis in bud, or can reduce the extent of injury of crisis to greatest extent.China in
Start within 2003《Public health emergency emergency reaction Mechanism for Building project》, by project implementation, country establishes
Communicable disease information reports management system (abbreviation reporting directly through network) based on case report,
Major infectious diseases monitoring automatic early-warning Information System configuration and pilots project were started in 2005 again.Thus, by near
The effort of 10 years, China sets up the monitoring of infectious disease automatic early-warning information system of the covering most crowds in the world.The system is used
The when Nuisance alarms method that dynamic percentile control chart method and spacescan statistic are combined, detects the exception of disease on one's own initiative
Assembled distribution.This method is that the aggregation of the case to having produced is detected, without perspective.And merely with
Total cases and incidence of disease information, other Surveillance species (meteorology, movement of population, article flows are not considered comprehensively
With purchase and sale, the medical illness of emphasis place and key population etc.), easily cause the imprecise, not in time of early warning.Equally, early warning threshold
The determination of value is also a problem.The complexity that other algorithm is calculated, large-scale infectious disease time-place clustering is carried out using model
Detection needs take a substantial amount of time and computing resource.
The content of the invention
The technology of the present invention solves problem:The present invention take using the currently monitored region certain specific infectious disease pathology,
Environmental risk key element predicts following a period of time inner region infectious disease severe, dead occurrence risk, thus carries out early warning, can
Overcome the problem of modes of warning of existing monitoring of infectious disease automatic early-warning information system does not have perspective.And the present invention is root
Severe, the different emergencies of dead occurrence risk are infected according to following a period of time, pre-warning signal is targetedly sent, can have
Effect solve tradition based on infectious disease spatial abnormal feature judge modes of warning judgement it is imprecise, not in time, be also easy to produce excessive early warning
The problems such as;Carried out available for the infectious disease early stage space-time modes of warning to the national monitoring of infectious disease automatic early-warning information platform of China
Supplement and optimization.
The present invention technical solution be:A kind of infectious disease severe based on viral recall rate, mortality risk early warning system
System and method, it makes full use of the specific infectious disease severe in the monitored area of current slot, dead related pathology, environmental key-element
To carry out infectious disease severe, the mortality risk early warning of next period, include as shown in Figure 1:
Risk definition module:The monitoring of infectious disease network data base of monitored area is scanned, chooses and passes by one in monitoring network
Each month and specific infectious disease average case and severe, death in historical time section patch, utilize shifting in the section time
Move percentile method to determine infectious disease severe, the historical baseline of incidence of mortality, thus define monitored area in the past one
Section the time in each moon the infectious disease severe, mortality risk grade.Risk class is defined into result deposit risk sample data
In storehouse.
Risk elements choose module:After risk definition module determines risk sample, risk elements are chosen module and adopted
With the means such as expert graded or reading document, analyzed in the numerous environmental key-elements in monitored area, choose the past with the period
The pathology of severe, mortality risk to infectious disease every month related previous moon, environmental key-element, set up risk elements body
System.Correlation analysis is recycled to determine that pathology, environment needed for final carry out infectious disease severe, mortality risk prediction will
Element.These key elements are subjected to sliding-model control, environmental key-element sample database is as a result stored in.While related diseases also to the current moon
Reason and environmental key-element data carry out same processing, are also deposited into environmental key-element sample database.
Risk network struction module:With reference to the data in risk sample database and environmental key-element sample database, utilize
Network structure learning method based on information theory, pathology, environment can most be reflected by setting up the monitored area in the past period
The Bayesian network of relation between key element and severe, dead occurrence risk;Node in the network set up represents finger respectively
Show infectious disease severe, dead occurrence risk grade variables and related pathologies, environmental key-element variable.The risk network that the module is set up
The severe that the currently monitored region infectious disease is carried out in risk assessment prediction module, mortality risk estimation will be pushed to.
Risk assessment prediction module:Risk assessment prediction module was read in environmental key-element sample database selected by the current moon
Related pathologies, environmental key-element data, the risk network set up using risk network struction module pass through joint probability formula meter
Calculate next month monitored area infectious disease severe, the possibility size of dead various risk class generations.On the basis of this, risk
The maximum risk class of assessment prediction module selection possibility is as next month severe, mortality risk grade, as judging bar
Part is pushed to grading forewarning system module.
Grading forewarning system module:According to risk assessment prediction module push next month monitored area the infection it is seriously ill
Disease, mortality risk grade, targetedly send two kinds of urgency level early warning or without early warning.Pre-warning signal is including risk etc.
Level, early warning are related to region and time segment information.Pre-warning signal can be with short message and network page form, to prevention and control of diseases portion
The full-time worker of door is pushed.
The risk network struction module implementation process is as follows:
(1) the Bayesian network initial configuration G=expressed with patterned way is first set up<V,E>, wherein structure chart G
In node V corresponding to be to indicate infectious disease severe, dead occurrence risk grade variables and the pathology selected, environmental key-element
Variable.Do not connected temporarily between node V between side E, sign variable temporarily without dependence;
(2) on the basis of monitored area pathology, environmental risk and infectious disease severe, dead occurrence risk historical summary,
Interactive information I (the v between each pair node are calculated using formula (1)i,vj)。
Pr(vi,vj) represent node to vi,vjThe simultaneous possibility size of situation shown in two kinds of representative variables,
Pr(vi)、Pr(vj) individual node v is represented respectivelyiOr vjThe simultaneous possibility size of situation shown in representative variable.It is right
It is more than the node pair of threshold epsilon in all interactive information, is sequentially inserted into from big to small in chained list L according to interactive information value size, this
Mono- pointer P of Shi Jianli points to chained list L first node pair.
(3) the first two node pair is removed from chained list L, is connected corresponding when being put into connection in E, and pointer P is moved
To next node pair.The node pair pointed by pointer P is taken out from chained list L, if there is no access path between the two nodes
In the presence of, just by it is corresponding while be added to connection while E in and by corresponding node to from chained list L delete, then pointer P is pointed to
Next node pair in chained list L.The step of edged is deleted is repeated, until pointer P is pointed in chained list L tail, or figure G
Untill when containing h-1 bar sides.H is the nodes in figure G.When figure G includes h-1 bars side, if adding a line to be formed back again
Road.
(4) pointer P is pointed to chained list L first node pair again, the node pointed by pointer P is taken out to vi,vj, in knot
The cut-off node set that can separate this to node contacts is found in composition G.It is single that each node centering is calculated using formula (2)
Node vnWith other a certain node v in structure chart GsBetween conditional mutual information I (vn,vs| C), C is set of node, vnFor node pair
vi,vjIn any one node.
Pr(vn,vs| C) represent set of node C interior joints to vi,vjSituation shown in two kinds of representative variables is simultaneous
Possibility size, Pr (vn|C)、Pr(vs| C) individual node v in set of node C is represented respectivelynOr vsFeelings shown in representative variable
The simultaneous possibility size of condition.If I (vn,vs| C) be less than threshold epsilon, then node is to vn,vsConditional sampling is unable to save in cut-off
Point set, then deletion of node is to v in connection side En,vsBetween connection side, increase node pair and cut-off node between connection
Side.The step is repeated, until pointer P points to chained list L tail.
(5) for the while E (v in connection in Ei,vj), if in node viOr vjBetween also there are other sides, then
By side E (vi,vj) temporarily deleted from connection side E.If the two nodes can not be spaced, by side E (vi,vj) rejoin
Connect side E;Otherwise permanent delet this edge.The side orientation finally given in current figure.
(6) when bayesian network structure once it is determined that, based on historical data data, in one section of past calculating network structure
In time under the pathology of the previous moon, environmental key-element combination condition, the various infectious disease severes of next month, dead occurrence risk
The joint probability distribution that grade occurs, thus obtains various infectious disease severes in the next month of monitored area, dead occurrence risk
The possibility size that grade occurs.Network structure can be optimized and revised according to historical summary.
A kind of infectious disease severe based on viral recall rate, mortality risk method for early warning, it is characterised in that step is as follows:
(1) according in the past period in the monitoring of infectious disease network of monitored area every month sum earlier month go through
Specific infectious disease average case and severe, death in history period patch, by mobile percentile method, observation is pressed
It is cycle severe, the incidence of mortality level that chronomere moves according to the moon, and synchronizing moving calculates history severe, incidence of mortality
Thus the percentile of level defines the weight of monitored area infectious disease every month in the past period as historical baseline
Disease, mortality risk grade.
(2) using expert graded or the reading means such as document, specific infectious disease severe, dead occurrence risk are set up
Pathology and environmental risk factors system, Causative virus recall rate, meteorological bar in the previous moon that factors system is related to risk assessment
Part and risk population are distributed multiple key elements.Then final carry out infectious disease severe is determined according to statistic correlation analysis, it is dead
Die the pathology and environmental risk key element needed for risk profile;Because Bayesian Network Inference input data is type or number of degrees
According to, so before evaluation of risk network is set up, with method of equal intervals, etc. frequency method or based on entropy method by risk elements data carry out from
Dispersion processing, automatically determines risk elements data from continuous type attribute to the corresponding relation of discrete type attribute.Method of equal intervals is basis
The interval number specified, by the codomain of numerical attribute, several are interval, and make each interval width equal.Deng frequency method and method of equal intervals
It is roughly the same, the difference is that method of equal intervals is to carry out equidistant divide to the frequency that data occur.And the method based on entropy is a kind of
Supervision, top-down discrete method.For discrete values attribute, there is the value of minimum entropy in this method selection numerical attribute
As split point, and recursively, division result is interval, obtains being layered discretization results.
(3) utilize the network structure learning method based on information theory, using monitored area in the past period it is each
Severe, the pathology of mortality risk and the previous moon, the environmental key-element data of the moon infectious disease, set up most reflecting for monitored area
The Bayesian network of relation between pathology, environmental key-element and severe, dead occurrence risk;Node difference in the network set up
Represent instruction infectious disease severe, dead occurrence risk grade variables and related pathologies, environmental key-element variable.
(4) risk network structure once it is determined that, on the basis of the pathology, environmental risk data in the monitored area current moon,
Infectious disease severe in the Bayesian network set up, the probability distribution of dead occurrence risk node are calculated, next month is thus obtained
The possibility size that the monitored area infectious disease severe, dead various risk class occur, the maximum risk of selection probability of happening
The risk that grade may occur the most as next month.
(5) according to the risk class of next month, the infectious disease severe, dead of monitored area in next month is thus judged
Risk is died likely in following three kinds of situations:It is (the infectious disease severe of next month, dead positioned at the high-order level of historical baseline
Incidence is died higher than period of history infectious disease severe of the same race, 80% value of the peak of incidence of mortality, i.e., >=P80, P is represented
Percentage numerical digit);Positioned at historical baseline by-level, (incidence of next month is located at the peak of period of history incidence
80% and 50% value is interval, i.e. P80-P50);Positioned at historical baseline reduced levels (when the incidence of next month is less than history
Phase average originating rate, i.e. ,≤P50);Then it is urgent according to next month monitored area infectious disease severe, the difference of mortality risk
Situation, targetedly sends early warning.When it is determined that severe, mortality risk are located at historical baseline high position level, control and prevention of disease people
Member should take most stringent prevention and control measure, such as isolation.And determine severe, mortality risk and be located at historical baseline by-level
When, control and prevention of disease personnel will consider prevention and control measure in two kinds of situation.It is a kind of it may be the case that epidemic situation has occurred and that, in gradually disappearing
Except the stage, then illustrate that early stage prevention and control measure is had an effect, control and prevention of disease personnel should be continuing with these measures;It is another
It may be the case that epidemic situation has not occurred, in the amount of the accumulateing stage, then control and prevention of disease personnel should be watched with the deepest concern, and this area is such a to be infected
The development of disease, it is necessary to can be targetedly to the general knowledge of public awareness campaign infectious disease prevention.
The advantage of the present invention compared with prior art is:The present invention can utilize disease severe and dead relevant risk key element
To predict infectious disease severe and dead occurrence risk, early warning is thus carried out, it is forward-looking;And can be to risk profile result
There are analysis of uncertainty, more science;It additionally is able to according to following a period of time infectious disease severe, dead occurrence risk size,
Classification does not carry out early warning, and the edidemic control personnel progress epidemic situation to Disease Prevention and Control Institutions and administrative department of public health is sentenced
Decision basis is provided in terms of disconnected, prevention and control measure selection.And the invention is established after network reasoning, once if missing dry-air blast
In the case of strategically located and difficult of access element can also forecasting risk, and constantly network can be adjusted according to expertise, this meets very much
The real work of infectious disease early warning needs.The present invention can be used as state suitable for the sick early warning of common transmittable in local cell domain
One supplement of family's monitoring of infectious disease automatic early-warning information platform infectious disease early stage space-time modes of warning, can improve China existing
Key area infectious disease early warning pattern.
Brief description of the drawings
Fig. 1 is a kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method for the present invention
Flow chart.
Fig. 2 is the Bayes for assessing 6 cities and counties of Hunan Province 4-7 month hand-foot-and-mouth diseases severe in 2013, death occurrence risk
Network structure
Embodiment
Below with " cities and counties of Hunan Province 6 (Changde City, Xiangxi Autonomous Prefecture, Xiangtan City, Loudi City, Huaihua City and Chenzhou City)
4-7 month hand-foot-and-mouth diseases severe, dead occurrence risk prediction and warning in 2013 " is instantiation, and specific introduce is based in the present invention
The specific implementation step (as shown in Figure 1) of the infectious disease severe of viral recall rate, mortality risk early warning system and method:
As shown in figure 1, a kind of infectious disease severe based on viral recall rate of the present invention, mortality risk early warning system and method
Module, risk network struction module, risk assessment prediction module, grading forewarning system mould are chosen by risk definition module, risk elements
Block is constituted.
Whole implementation process is as follows:
(1) the monitoring of infectious disease network data base of 6 cities and counties is scanned, chooses each in the 4-7 months in 2012 in monitoring network
Hand-foot-and-mouth disease average case in individual month and historical time section patch (the 2010-2011 months and its front and rear each two months) with
Severe, death, the salty disease of brothers, the historical baseline of incidence of mortality are determined using mobile percentile method, thus
Define 6 cities and counties the infectious disease severe, mortality risk grade in the 4-7 months in 2012.Risk class is defined into result deposit risk
In sample database.
(2) after risk definition module determines risk sample, risk elements choose module using expert graded or
Read the means such as document, analyze, be chosen in the 3-6 months in 2012 and hand-foot-and-mouth disease every month in the numerous environmental key-elements of 6 cities and counties
Severe, the related pathology of mortality risk, environmental key-element, set up risk elements system.Correlation analysis is recycled to determine most
The required pathology of the carry out infectious disease severe at end, mortality risk prediction, environmental key-element.These key elements are subjected to sliding-model control,
As a result it is stored in environmental key-element sample database.While related pathologies and environmental key-element number also to every month in the 3-6 months in 2013
According to same processing is carried out, also it is deposited into environmental key-element sample database.
(3) data in risk sample database and environmental key-element sample database are combined, the net based on information theory is utilized
Network Structure learning method, set up in the past period 6 cities and counties can most reflect pathology, environmental key-element and severe, dead hair
The Bayesian network of relation between raw risk;Node in the network set up represents instruction hand-foot-and-mouth disease severe respectively, dead
Die occurrence risk grade variables and related pathologies, environmental key-element variable.The risk network that the module is set up will be pushed to risk and comment
Estimate in prediction module carry out 4-7 months in 2013 in every month 6 cities and counties' hand-foot-and-mouth disease severe, mortality risk estimate.
(4) risk assessment prediction module reads related diseases selected in the 3-6 months in 2013 in environmental key-element sample database
Reason, environmental key-element data, the risk network set up using risk network struction module calculate 2013 by joint probability formula
In year 4-7 months every month 6 cities and counties' hand-foot-and-mouth disease severe, the possibility size of dead various risk class generations.On the basis of this,
The maximum risk class of risk assessment prediction module selection possibility is used as the severe of every month, dead wind in the 4-7 months in 2013
Dangerous grade, grading forewarning system module is pushed to as Rule of judgment.
(5) according to risk assessment prediction module push 4-7 months in 2013 in every month often cities and counties the infectious disease severe,
Mortality risk grade, targetedly sends two kinds of urgency level early warning or without early warning.Pre-warning signal include risk class,
Early warning is related to region and time segment information.Pre-warning signal can be with short message and network page form, to prevention and control of diseases department
Full-time worker pushed.
Above-mentioned each module to implement process as follows:
1. risk definition module
According to each moon and same period the past (2010-2012 in the 4-7 months in 2012 in the monitoring of infectious disease network of 6 cities and counties
Year month and its front and rear each two months) historical time section patch in hand-foot-and-mouth disease average case and severe, death,
By mobile percentile method, observe according to the cycle severe, incidence of mortality level that the moon is chronomere's movement, and synchronously
Mobile computing history severe, incidence of mortality horizontal base line (percentile), thus define 6 cities and counties' hand-foot-and-mouth disease severes, death
Risk.≤ P50 (incidence of each month is less than period of history average originating rate, and P represents percentage numerical digit) is wind force 0 danger,
P80-P50 (80% and 50% value that the incidence of each month is located at the peak of period of history incidence is interval) is 1 grade
Risk, >=P80 (higher than period of history infectious disease severe of the same race, death send out by the infectious disease severe of each month, incidence of mortality
80% value of the peak of raw rate) it is gentle breeze danger.Risk class is defined in result deposit risk sample database.
2. risk elements choose module
Using the means such as expert graded or reading document, hand-foot-and-mouth disease severe, the pathology of dead occurrence risk are set up
With environmental risk factors system, factors system is related to severe, mortality risk phase in the 3-6 months in 2012 with hand-foot-and-mouth disease every month
The viral total recall rates of the EV71 for less than the 5 years old children closed, each age group (1:Less than 6 months, 2:- 1 year old 6 months,:3:1-2
Year, 4:2-4 Sui, 5:4-5 Sui, 6:More than 5 years old) EV71 viral recall rate, hand-foot-and-mouth disease severe, incidence of mortality, the current moon
Meteorological condition (minimum relative humidity, average relative humidity, sunshine duration, the lowest temperature, temperature on average, the most highest temperature, strong wind
Speed, mean wind speed, average gas pressure, rainfall), multiple key elements such as Susceptible population's (less than 5 years old) density of population and the level of urbanization.
The pathology needed for final carry out hand-foot-and-mouth disease severe, mortality risk prediction is determined according to statistic correlation analysis
With environmental risk key element.The risk elements of Bayesian network method are finally entered including every with hand-foot-and-mouth disease in the 3-6 months in 2012
The total recall rate of the severe of individual month, the EV71 viruses of less than 5 years old children of mortality risk correlation, -1 year old 6 months, 1-2 Sui, 4-5
The viral recall rates of the children EV71 in year and more than 5 years old, hand-foot-and-mouth disease severe, incidence of mortality, current moon meteorological condition (minimum phase
To humidity, average relative humidity, sunshine duration, the lowest temperature, temperature on average, the highest temperature, maximum wind velocity, mean wind speed, flat
Equal air pressure, rainfall), Susceptible population's (less than 5 years old) density of population and the level of urbanization.With based on entropy method by risk elements number
According to sliding-model control is carried out, risk elements data are automatically determined from continuous type attribute to the corresponding relation of discrete type attribute.It is based on
Division result is interval as split point, and recursively for the value with minimum entropy in the method choice numerical attribute of entropy, so as to obtain
It is layered discretization results.Related pathologies and environmental key-element data also simultaneously to every month in the 3-6 months in 2013 carry out same
Processing.Two groups of pathology, environmental key-element data are all deposited into environmental key-element sample database.
3. risk network struction module
With reference to the data in risk sample database and environmental key-element sample database, the network knot based on information theory is utilized
Structure learning method, sets up most reflecting pathology, environmental key-element and severe, dead occurring wind in the past period 6 cities and counties
The Bayesian network of relation between danger;Node in the network set up represents instruction hand-foot-and-mouth disease severe, dead hair respectively
Raw risk grade variables and related pathologies, environmental key-element variable.Step is as follows:
1. the Bayesian network initial configuration G=expressed with patterned way is first set up<V,E>, wherein structure chart G
In node V corresponding to be to indicate that hand-foot-and-mouth disease severe, dead occurrence risk grade variables and the pathology selected, environment will
Plain variable.Do not connected temporarily between node V between side E, sign variable temporarily without dependence;
2. on the basis of 6 cities and counties' pathology, environmental risk and hand-foot-and-mouth disease severe, dead occurrence risk historical summary, profit
Interactive information I (the v between each pair node are calculated with formula (1)i,vj)。
Pr(vi,vj) represent node to vi,vjThe simultaneous possibility size of situation shown in two kinds of representative variables,
Pr(vi)、Pr(vj) individual node v is represented respectivelyiOr vjThe simultaneous possibility size of situation shown in representative variable.It is right
It is more than 0.5 node pair in all interactive information, is sequentially inserted into from big to small in chained list L according to interactive information value size, now
Set up first node pair that a pointer P points to chained list L.
3. the first two node pair is removed from chained list L, is connected corresponding when being put into connection in E, and pointer P is moved on to
Next node pair.The node pair pointed by pointer P is taken out from chained list L, if no access path is deposited between the two nodes
, just by it is corresponding while be added to connection while E in and by corresponding node to from chained list L delete, then by pointer P point to chain
Next node pair in table L.The step of edged is deleted is repeated, is wrapped until pointer P is pointed in chained list L tail, or figure G
Untill when having contained h-1 bar sides.H is the nodes in figure G.When figure G includes h-1 bars side, if adding a line to form loop again.
4. pointer P is pointed to chained list L first node pair again, the node pointed by pointer P is taken out to vi,vj, in structure
The cut-off node set that can separate this to node contacts is found in figure G.The single section of each node centering is calculated using formula (2)
Point vnWith other a certain node v in structure chart GsBetween conditional mutual information I (vn,vs| C), C is set of node, vnIt is node to vi,
vjIn any one node.
Pr(vn,vs| C) represent set of node C interior joints to vi,vjSituation shown in two kinds of representative variables is simultaneous
Possibility size, Pr (vn|C)、Pr(vs| C) individual node v in set of node C is represented respectivelynOr vsFeelings shown in representative variable
The simultaneous possibility size of condition.If I (vn,vs| C) be less than threshold epsilon, then node is to vn,vsConditional sampling is unable to save in cut-off
Point set, then deletion of node is to v in connection side En,vsBetween connection side, increase node pair and cut-off node between connection
Side.The step is repeated, until pointer P points to chained list L tail.
5. for the while E (v in connection in Ei,vj), if in node viOr vjBetween also there are other sides, then
By side E (vi,vj) temporarily deleted from connection side E.If the two nodes can not be spaced, by side E (vi,vj) rejoin
Connect side E;Otherwise permanent delet this edge.The side orientation finally given in current figure.
6. when bayesian network structure once it is determined that, pathology, environment based on each cities and counties 3-6 every months month in 2012 will
Wind occurs for 3 kinds of hand-foot-and-mouth disease severes, death under various pathology, environmental key-element combination condition in prime number evidence, calculating network structure
The joint probability distribution that dangerous grade occurs, thus obtains 3 kinds of each cities and counties every month in the 4-7 months in 2012 of corresponding 6 cities and counties
The possibility size that hand-foot-and-mouth disease severe, dead occurrence risk grade occur.Shellfish is verified and adjusted using true Monitoring Data
This network structure of leaf.
4. risk assessment prediction module
Related pathologies selected in the 3-6 months in 2013 in environmental key-element sample database, environmental key-element data are read, are utilized
The risk network that risk network struction module is set up, cities and counties' every month 6 in the 4-7 months in 2013 are calculated by joint probability formula
The possibility size that hand-foot-and-mouth disease severe, dead various risk class occur.On the basis of this, the selection of risk assessment prediction module
The maximum risk class of possibility is pushed away as severe, the mortality risk grade of every month in the 4-7 months in 2013 as Rule of judgment
Give grading forewarning system module.
5. grading forewarning system module
The risk class result come is pushed according to risk assessment prediction module, thus judges 6 cities and counties the 4-7 months in 2013
In hand-foot-and-mouth disease severe in every month each cities and counties' next month, mortality risk likely in following three kinds of situations:Position
In the high-order level of historical baseline, (the infectious disease severe of next month, incidence of mortality are seriously ill higher than period of history infection of the same race
Disease, 80% value of the peak of incidence of mortality, i.e., >=P80, P represent percentage numerical digit);Positioned at historical baseline by-level
(80% and 50% value that the incidence of next month is located at the peak of period of history incidence is interval, i.e. P80-P50);Position
In historical baseline reduced levels (incidence of next month is less than period of history average originating rate, i.e. ,≤P50);Then according to every
Hand-foot-and-mouth disease severe, the different emergencies of mortality risk in individual month each cities and counties' next month, targetedly send early warning.
When it is determined that severe, mortality risk are located at historical baseline high position level, control and prevention of disease personnel should take most stringent prevention and control to arrange
Apply, such as isolation.And when determining that severe, mortality risk are located at historical baseline by-level, control and prevention of disease personnel will be divided to two kinds
Situation considers prevention and control measure.It is a kind of it may be the case that epidemic situation has occurred and that, in gradually eliminating the stage, then illustrate early stage prevention and control
Measure is had an effect, and control and prevention of disease personnel should be continuing with these measures;It is another it may be the case that epidemic situation has not occurred,
In the amount of the accumulateing stage, then control and prevention of disease personnel should watch with the deepest concern the development of such a infectious disease in this area, it is necessary to can be directed to
Property to public awareness campaign infectious disease prevention general knowledge.
The content not being described in detail in description of the invention belongs to prior art known to professional and technical personnel in the field.
Claims (2)
1. a kind of infectious disease severe based on viral recall rate, mortality risk early warning system, it is characterised in that including:Risk is defined
Module, risk elements choose module, risk network struction module, risk assessment prediction module and grading forewarning system module;Wherein:
Risk definition module:When passing by one section in the monitoring of infectious disease network data base of scanning monitored area, selection monitoring network
Between in each month and specific infectious disease average case and severe, death in historical time section patch, utilize mobile hundred
Quantile method determines infectious disease severe, the historical baseline of incidence of mortality, thus defines monitored area at one section of past
Between in each moon the infectious disease severe, mortality risk grade;Risk class is defined in result deposit risk sample database;
Risk elements choose module:After risk definition module determines risk sample, risk elements choose module using special
Family scoring or reading document means, in the numerous environmental key-elements in monitored area analyze, choose the past with the period in the biography
Severe, the pathology of the previous moon of mortality risk correlation, the environmental key-element caught an illness every month, set up risk elements system;It is sharp again
The final pathology carried out needed for infectious disease severe, mortality risk are predicted, environmental key-element are determined with correlation analysis;By these
Key element carries out sliding-model control, is as a result stored in environmental key-element sample database;While related pathologies and environment also to the current moon
Factor data carries out same processing, is also deposited into environmental key-element sample database;
Risk network struction module:With reference to the data in risk sample database and environmental key-element sample database, using based on
The network structure learning method of information theory, pathology, environmental key-element can most be reflected by setting up the monitored area in the past period
The Bayesian network of relation between severe, dead occurrence risk;Node in the network set up represents instruction and passed respectively
Catch an illness severe, dead occurrence risk grade variables and related pathologies, environmental key-element variable;The risk network that the module is set up will be pushed away
It is sent to the severe that the current moon monitored area infectious disease is carried out in risk assessment prediction module, mortality risk estimation;
Risk assessment prediction module:Risk assessment prediction module reads the correlation selected by the current moon in environmental key-element sample database
Pathology, environmental key-element data, the risk network set up using risk network struction module, under being calculated by joint probability formula
The possibility size that one month monitored area infectious disease severe, dead various risk class occur;On the basis of this, risk assessment
The maximum risk class of prediction module selection possibility is pushed away as next month severe, mortality risk grade as Rule of judgment
Give grading forewarning system module;
Grading forewarning system module:Pushed according to risk assessment prediction module next month monitored area the infectious disease severe, dead
Risk class is died, two kinds of urgency level early warning are targetedly sent or without early warning;Pre-warning signal includes risk class, pre-
Police is related to region and time segment information;Pre-warning signal is with short message and network page form, to the sole duty of prevention and control of diseases department
Staff is pushed;
Described risk network struction module implementation process is as follows:
(1) the Bayesian network initial configuration G=expressed with patterned way is first set up<V,E>, in wherein structure chart G
Corresponding to node V is to indicate that infectious disease severe, dead occurrence risk grade variables and the pathology selected, environmental key-element become
Amount;Do not connected temporarily between node V between side E, sign variable temporarily without dependence;
(2) on the basis of monitored area pathology, environmental risk and infectious disease severe, dead occurrence risk historical summary, utilize
Formula (1) calculates the interactive information I (v between each pair nodei,vj);
Pr(vi,vj) represent node to vi,vjThe simultaneous possibility size of situation, Pr shown in two kinds of representative variables
(vi)、Pr(vj) individual node v is represented respectivelyiOr vjThe possibility size that situation shown in representative variable occurs respectively;For
All interactive information are more than the node pair of threshold epsilon, are sequentially inserted into from big to small in chained list L according to interactive information value size, now
Set up first node pair that a pointer P points to chained list L;
(3) the first two node pair is removed from chained list L, is connected corresponding when being put into connection in E, and pointer P is moved on to down
One node pair;The node pair pointed by pointer P is taken out from chained list L, if there is no access path presence between the two nodes,
Just by it is corresponding when being added to connection in E and by corresponding node to being deleted from chained list L, pointer P is then pointed into chained list L
In next node pair;The step of edged is deleted is repeated, is included until pointer P is pointed in chained list L tail, or figure G
Untill during h-1 bar sides;H is the nodes in figure G;When figure G includes h-1 bars side, if adding a line to form loop again;
(4) pointer P is pointed to chained list L first node pair again, the node pointed by pointer P is taken out to vi,vj, in structure chart G
Middle searching can separate this cut-off node set to node contacts;Each node centering individual node is calculated using formula (2)
vnWith other a certain node v in structure chart GsBetween conditional mutual information I (vn,vs| C), C is set of node, vnIt is node to vi,vj
In any one node;
Pr(vn,vs| C) represent set of node C interior joints to vi,vjThe simultaneous possibility of situation shown in two kinds of representative variables
Size, Pr (vn|C)、Pr(vs| C) individual node v in set of node C is represented respectivelynOr vsSituation shown in representative variable occurs
Possibility size;If I (vn,vs| C) be less than threshold epsilon, then node is to vn,vsConditional sampling is unable in cut-off set of node, then
Deletion of node is to v in connection side En,vsBetween connection side, increase node pair and cut-off node between connection side;Repeat
The step, until pointer P points to chained list L tail;
(5) for the while E (v in connection in Ei,vj), if in node viOr vjBetween also there are other sides, then by side E
(vi,vj) temporarily deleted from connection side E, if the two nodes can not be spaced, by side E (vi,vj) rejoin connection side
E;Otherwise permanent delet this edge;The side orientation finally given in current figure;
(6) when bayesian network structure once it is determined that, based on historical data data, in the past period calculating network structure
In under the pathology of the previous moon, environmental key-element combination condition, the various infectious disease severes of next month, dead occurrence risk grade
The joint probability distribution of generation, thus obtains various infectious disease severes in the next month of monitored area, dead occurrence risk grade
The possibility size of generation;Network structure is optimized and revised according to historical summary.
2. a kind of infectious disease severe based on viral recall rate, mortality risk method for early warning, it is characterised in that step is as follows:
(1) according in the past period in the monitoring of infectious disease network of monitored area every month sum earlier month history when
Between specific infectious disease average case in section patch and severe, death, by mobile percentile method, observe according to the moon
Cycle severe, the incidence of mortality level moved for chronomere, and synchronizing moving calculates history severe, incidence of mortality level
Percentile as historical baseline, thus define monitored area in the past period the severe of infectious disease every month,
Mortality risk grade;
(2) using expert graded or reading document means, set up specific infectious disease severe, the pathology of dead occurrence risk and
Environmental risk factors system, Causative virus recall rate, meteorological condition and wind in the previous moon that factors system is related to risk assessment
The dangerous multiple key elements of Crowds Distribute;Then final carry out infectious disease severe, mortality risk are determined according to statistic correlation analysis
Pathology and environmental risk key element needed for prediction;Due to Bayesian Network Inference input data be type or level data, so
Before evaluation of risk network is set up, with method of equal intervals, etc. frequency method or based on entropy method risk elements data are carried out at discretization
Reason, automatically determines risk elements data from continuous type attribute to the corresponding relation of discrete type attribute;
(3) the network structure learning method based on information theory is utilized, every month should in the past period using monitored area
Severe, the pathology of mortality risk and the previous moon, the environmental key-element data of infectious disease, that sets up monitored area can most reflect disease
The Bayesian network of relation between reason, environmental key-element and severe, dead occurrence risk;Node in the network set up generation respectively
Table instruction infectious disease severe, dead occurrence risk grade variables and related pathologies, environmental key-element variable;
(4) risk network structure once it is determined that, on the basis of the pathology, environmental risk data in the monitored area current moon, calculate
Infectious disease severe, the probability distribution of dead occurrence risk node, thus obtain next month monitoring in the Bayesian network of foundation
The possibility size that the region infectious disease severe, dead various risk class occur, the maximum risk class of selection probability of happening
The risk that may occur the most as next month;
(5) according to the risk class of next month, the infectious disease severe in next month of monitored area, dead wind are thus judged
Danger is likely in following three kinds of situations:Positioned at the high-order level of historical baseline, historical baseline high position level refers to next month
The infectious disease severe, incidence of mortality be higher than period of history infectious disease severe of the same race, the 80% of the peak of incidence of mortality
Value, i.e., >=P80, P represents percentage numerical digit;Positioned at historical baseline by-level, historical baseline by-level refers to next month
The incidence peak that is located at period of history incidence 80% and 50% value it is interval, i.e. P80-P50;Positioned at historical baseline
Reduced levels, historical baseline reduced levels refer to that the incidence of next month is less than period of history average originating rate, i.e. ,≤P50;
Then according to next month monitored area infectious disease severe, the different emergencies of mortality risk, targetedly send pre-
It is alert;When it is determined that severe, mortality risk are located at historical baseline high position level, control and prevention of disease personnel should take most stringent prevention and control
Measure;And when determining that severe, mortality risk are located at historical baseline by-level, control and prevention of disease personnel will consider to prevent in two kinds of situation
Control measure;It is a kind of it may be the case that epidemic situation has occurred and that, in the stage is gradually eliminated, then illustrate that early stage prevention and control measure occurs
Effect, control and prevention of disease personnel should be continuing with these measures;It is another it may be the case that epidemic situation has not occurred, in fermenting rank
Section, then control and prevention of disease personnel should watch with the deepest concern the development of such a infectious disease in this area, targetedly to public awareness campaign infectious disease
The general knowledge of protection;
That monitored area is set up in the step (3) can most reflect pass between pathology, environmental key-element and severe, dead occurrence risk
The Bayesian network of system to implement process as follows:
(31) the Bayesian network initial configuration G=expressed with patterned way is first set up<V,E>, in wherein structure chart G
Node V corresponding to be to indicate that infectious disease severe, dead occurrence risk grade variables and the pathology selected, environmental key-element become
Amount;Do not connected temporarily between node V between side E, sign variable temporarily without dependence;
(32) on the basis of monitored area pathology, environmental risk and infectious disease severe, dead occurrence risk historical summary, utilize
Formula (1) calculates the interactive information I (v between each pair nodei,vj);
Pr(vi,vj) represent node to vi,vjThe simultaneous possibility size of situation, Pr shown in two kinds of representative variables
(vi)、Pr(vj) individual node v is represented respectivelyiOr vjThe possibility size that situation shown in representative variable occurs;For all
Interactive information is more than the node pair of threshold epsilon, is sequentially inserted into chained list L, is now set up from big to small according to interactive information value size
One pointer P points to chained list L first node pair;
(33) the first two node pair is removed from chained list L, is connected corresponding when being put into connection in E, and pointer P is moved on to down
One node pair;The node pair pointed by pointer P is taken out from chained list L, if there is no access path presence between the two nodes,
Just by it is corresponding when being added to connection in E and by corresponding node to being deleted from chained list L, pointer P is then pointed into chained list L
In next node pair;The step of edged is deleted is repeated, is included until pointer P is pointed in chained list L tail, or figure G
Untill during h-1 bar sides;H is the nodes in figure G;When figure G includes h-1 bars side, if adding a line to form loop again;
(34) pointer P is pointed to chained list L first node pair again, the node pointed by pointer P is taken out to vi,vj, in structure chart
The cut-off node set that can separate this to node contacts is found in G;Each node centering individual node is calculated using formula (2)
vnWith the conditional mutual information I (v between other a certain node vs in structure chart Gn,vs| C), C is set of node, vnIt is node to vi,vj
In any one node;
Pr(vn,vs| C) represent set of node C interior joints to vi,vjThe simultaneous possibility of situation shown in two kinds of representative variables
Size, Pr (vn|C)、Pr(vs| C) individual node v in set of node C is represented respectivelynOr vsSituation shown in representative variable occurs
Possibility size;If I (vn,vs| C) be less than threshold epsilon, then node is to vn,vsConditional sampling is unable in cut-off set of node, then
Deletion of node is to v in connection side En,vsBetween connection side, increase node pair and cut-off node between connection side;Repeat
The step, until pointer P points to chained list L tail;
(35) for the while E (v in connection in Ei,vj), if in node viOr vjBetween also there are other sides, then by side
E(vi,vj) temporarily deleted from connection side E, if the two nodes can not be spaced, by side E (vi,vj) rejoin connection
Side E;Otherwise permanent delet this edge;The side orientation finally given in current figure;
(36) when bayesian network structure once it is determined that, based on historical data data, in calculating network structure at one section of past
Between under the pathology of the previous moon, environmental key-element combination condition, the various infectious disease severes of next month, dead occurrence risk etc.
The joint probability distribution that level occurs, thus obtains various infectious disease severes in the next month of monitored area, dead occurrence risk etc.
The possibility size that level occurs;Network structure is optimized and revised according to historical summary.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510056900.2A CN104615884B (en) | 2015-02-04 | 2015-02-04 | A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510056900.2A CN104615884B (en) | 2015-02-04 | 2015-02-04 | A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104615884A CN104615884A (en) | 2015-05-13 |
CN104615884B true CN104615884B (en) | 2017-07-28 |
Family
ID=53150325
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510056900.2A Active CN104615884B (en) | 2015-02-04 | 2015-02-04 | A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104615884B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107133901B (en) * | 2017-05-05 | 2020-11-03 | 中国科学院深圳先进技术研究院 | Infectious disease prevention and control method and system |
CN107680685A (en) * | 2017-10-24 | 2018-02-09 | 山东浪潮云服务信息科技有限公司 | A kind of disease pre-warning method and system |
CN109949930A (en) * | 2019-03-22 | 2019-06-28 | 武汉工程大学 | Animal epidemics diagnostic method and system based on Bayesian network |
CN110517770B (en) * | 2019-08-28 | 2021-08-17 | 中国科学院重庆绿色智能技术研究院 | Perperative critical adverse event intervention decision method based on Bayesian network and utility system |
CN112205965B (en) * | 2020-08-28 | 2022-02-22 | 北京大学 | Health risk key event detection method and system based on time window cutting |
CN112185583B (en) * | 2020-10-14 | 2022-05-31 | 天津之以科技有限公司 | Data mining quarantine method based on Bayesian network |
CN112365942B (en) * | 2020-10-20 | 2022-05-31 | 哈尔滨学院 | Epidemic risk prediction analysis method for infectious diseases |
CN112185586A (en) * | 2020-11-17 | 2021-01-05 | 北京嘉和海森健康科技有限公司 | Infectious disease monitoring and early warning method and device |
CN113159571A (en) * | 2021-04-20 | 2021-07-23 | 中国农业大学 | Cross-border foreign species risk level determination and intelligent identification method and system |
CN115810425B (en) * | 2022-11-30 | 2023-12-08 | 广州中医药大学第一附属医院 | Method and device for predicting mortality risk level of sepsis shock patient |
CN117133461B (en) * | 2023-10-23 | 2024-01-30 | 北京肿瘤医院(北京大学肿瘤医院) | Method and device for postoperative short-term death risk assessment of aged lung cancer patient |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101894309A (en) * | 2009-11-05 | 2010-11-24 | 南京医科大学 | Epidemic situation predicting and early warning method of infectious diseases |
CN202711253U (en) * | 2012-03-09 | 2013-01-30 | 李晓松 | Detecting and early-stage pre-warning system for infectious disease aggregation |
CN103259855A (en) * | 2013-04-27 | 2013-08-21 | 北京福乐云检测科技有限公司 | Cloud platform of infectious disease and chronic disease detecting and monitoring pre-warning system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104956389A (en) * | 2012-10-23 | 2015-09-30 | 赛拉诺斯股份有限公司 | Assisted medical and associated lifestyle decision making |
-
2015
- 2015-02-04 CN CN201510056900.2A patent/CN104615884B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101894309A (en) * | 2009-11-05 | 2010-11-24 | 南京医科大学 | Epidemic situation predicting and early warning method of infectious diseases |
CN202711253U (en) * | 2012-03-09 | 2013-01-30 | 李晓松 | Detecting and early-stage pre-warning system for infectious disease aggregation |
CN103259855A (en) * | 2013-04-27 | 2013-08-21 | 北京福乐云检测科技有限公司 | Cloud platform of infectious disease and chronic disease detecting and monitoring pre-warning system |
Non-Patent Citations (3)
Title |
---|
传染病疫情早期预警的主要模型;王小莉等;《现代预防医学》;20081125;第35卷(第22期);第4339-4341页 * |
基于不同参数的空间扫描统计量方法在传染病暴发探测中的效果比较;李小洲等;《中国卫生统计》;20140425;第31卷(第2期);第236-239页 * |
重点传染病预防控制综合分析系统的设计与实现;陈勇;《中国优秀硕士学位论文全文数据库》;20091115(第11期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN104615884A (en) | 2015-05-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104615884B (en) | A kind of infectious disease severe based on viral recall rate, mortality risk early warning system and method | |
Park et al. | Real-time prediction and avoidance of secondary crashes under unexpected traffic congestion | |
Chen et al. | Road traffic congestion monitoring in social media with hinge-loss Markov random fields | |
Park et al. | Interpretation of Bayesian neural networks for predicting the duration of detected incidents | |
Li et al. | Competing risk mixture model and text analysis for sequential incident duration prediction | |
CN110544402B (en) | Chemical industry park virtual simulation emergency deduction system based on big data fusion | |
Abdel-Aty et al. | Assessing safety on Dutch freeways with data from infrastructure-based intelligent transportation systems | |
CN106507315B (en) | Urban traffic accident prediction technique and system based on network social intercourse media data | |
Van Patten et al. | A microspatial analysis of robbery: Prospective hot spotting in a small city | |
CN108417274A (en) | Forecast of epiphytotics method, system and equipment | |
CN106330533A (en) | Real-time topology establishment method of large-scale network alarms | |
CN106295565A (en) | Monitor event identifications based on big data and in real time method of crime prediction | |
Yuan et al. | Modeling real-time cycle-level crash risk at signalized intersections based on high-resolution event-based data | |
Codeço et al. | Estimating the effective reproduction number of dengue considering temperature-dependent generation intervals | |
CN106447109B (en) | A kind of community policy patrol path optimization's design method based on deep learning algorithm | |
CN109118039B (en) | Inland water area oriented ship self-sinking important risk identification, prevention and control device and method | |
Pan et al. | Discovering spatial-temporal patterns via complex networks in investigating COVID-19 pandemic in the United States | |
Moore et al. | Local and regional dynamics of chikungunya virus transmission in Colombia: the role of mismatched spatial heterogeneity | |
CN110533480A (en) | A kind of market passenger flow forecasting based on big data | |
Yoo et al. | Machine learning‐driven dynamic risk prediction for highly pathogenic avian influenza at poultry farms in Republic of Korea: Daily risk estimation for individual premises | |
Yijing et al. | Intelligent algorithms for incident detection and management in smart transportation systems | |
CN114707719A (en) | Personnel behavior prediction method for dangerous area | |
Koch et al. | Predicting cannabis cultivation on national forests using a rational choice framework | |
Liu et al. | A real-time explainable traffic collision inference framework based on probabilistic graph theory | |
Paudel et al. | Factors that affect the timing of the dispatch of initial attack resources to forest fires in northeastern Ontario, Canada |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |