A Method for Matching the Supply and Demand of Telemedicine Considering the Doctor-Patient Relationship and Intermediary Intervention

Abstract

The present invention discloses a remote medical supply-demand matching method considering the doctor-patient relationship and intermediary intervention, belonging to the technical field of remote medical treatment. It includes heterogeneous information distance measurement, matching subject preference utility calculation, preference utility correction, measurement of the two-level demand mutual relationship considering the peer effect, construction and solution of a multi-objective optimization model, solves the technical problem of supply-demand matching in the remote medical treatment scenario, and starting from the characteristics and requirements of the remote medical service supply-demand matching decision problem, through analyzing the bounded rationality characteristics, psychological behaviors of the matching subjects and the interaction relationship between the two-level demanders, a supply-demand matching decision method considering the two-level demand network is given. It fully considers each participating subject and its preference in the matching decision process, simultaneously considers the certainty and fuzziness of the decision-maker's evaluation, as well as the influence degree of the doctor on the patient's decision in the doctor-patient interaction process, enhances the scientificity and feasibility of the decision-making, and makes up for the deficiencies of the current research.

Description

Technical Field

[0001] The present invention belongs to the technical field of telemedicine, and particularly relates to a telemedicine supply-demand matching method considering doctor-patient relationship and intermediary intervention. Background Art

[0002] Telemedicine realizes the connection between high-quality medical resources and medical needs across regions through telecommunication technologies, remoteizes medical services in a multi-organization cooperation and multi-agent collaboration manner, realizes the linkage of medical resources across regions, and improves the efficiency of medical information dissemination.

[0003] In the problem of telemedicine service supply-demand matching, there are usually three parties: patients, applicant doctors, and experts. Patients are the demanders of telemedicine services and urgently need high-quality medical resources; applicant doctors are both the demanders of telemedicine services, needing the guidance of superior experts, and the intermediary parties of telemedicine services, initiating telemedicine applications to connect patients and superior experts; experts are the providers of telemedicine services, providing high-quality medical services.

[0004] In recent years, telemedicine has formed a new medical service model of Doctor-to-Doctor-to-Patient (D2D2P), that is, patients have their first diagnosis in local medical institutions. If the attending doctor is unable to provide a clear diagnosis or treatment for the patient due to reasons such as insufficient doctor level and backward medical institution facilities; after obtaining the patient's consent, primary doctors can apply for corresponding telemedicine services to a superior cooperative hospital to seek the help of experts; after matching, the two doctors discuss the patient's condition at a scheduled time and reach a consistent plan, so that the patient can enjoy high-quality medical resources in large hospitals "without leaving home", keep the patient in the local area, and relieve the problems of expensive medical treatment and difficult access to medical services for the people.

[0005] Currently, the following problems exist in telemedicine:

[0006] Telemedicine has formed a two-level supply-demand network with multi-agent participation of D2D2P. Traditional supply-demand matching problems only consider the two parties of supply and demand, and do not fully consider each participating subject and its preferences in the matching decision-making process, as well as the mutual relationships among the subjects.

[0007] Due to the professionalism of medical knowledge, patients are more or less influenced by doctors in the medical decision-making process, and the current traditional technologies do not consider the effect of this influence on the matching result.

[0008] Only considering a single form of evaluation type, without constructing a matching decision model according to the evaluation preferences and habits of decision-makers and the certainty and ambiguity of decision-maker evaluations, the feasibility and rationality of the matching result are relatively weak. Summary of the Invention

[0009] The purpose of the present invention is to provide a telemedicine supply and demand matching method that takes into account the doctor-patient relationship and intermediary intervention, thereby solving the technical problem of supply and demand matching in a telemedicine context.

[0010] To achieve the above object, the present invention adopts the following technical solution: a telemedicine supply and demand matching method considering doctor-patient relationship and intermediary intervention, comprising the following steps:

[0011] Step 1: Collect heterogeneous information provided by the supply and demand parties participating in the telemedicine service through the telemedicine information service platform; the supply and demand parties include multiple matching subjects; the types of matching subjects include demand parties and supply parties, the demand parties are patients or applicant doctors, the supply parties are experts, the applicant doctors are the patients' attending doctors, and the applicant doctors and patients, experts and patients are matched using a one-to-many matching method;

[0012] Heterogeneous information includes the expectation information and evaluation information of each matching subject on its matching object; in the heterogeneous information, the mutual evaluation matrix of the matching subjects is constructed according to the evaluation information, and the expectation matrix of each matching subject is constructed according to the expectation information;

[0013] Step 2: Take the expected value of each matching subject as the reference point to calculate the distance measurement of heterogeneous information, and obtain the distance measurement result. The gap between the actual evaluation value and the expected value in the heterogeneous information is used as the condition to measure the preference utility of each matching subject.

[0014] Step 3: Based on the distance measurement results, establish the profit and loss matrix of each matching subject relative to the reference point, calculate the gains and losses of each matching subject based on the profit and loss matrix, and consider the different psychological states of the matching subject towards gains and losses, and establish the prospect decision matrix of each matching subject under the condition of limited rationality;

[0015] The preferred utility of the matching subject is represented by integrating the prospect values in the prospect decision matrix;

[0016] Step 4: Use the disappointment theory to measure each matching subject's psychological perception of the matching results, so as to modify the preference utility of each matching subject;

[0017] Step 5: Based on grey correlation analysis, measure the influence of different doctors on patients’ decision making, take patients’ expectations of experts as reference sequence, and improve the application of grey correlation function in heterogeneous information according to the distance between the expected information of doctors and the reference sequence;

[0018] Step 6: Based on the methods of steps 1 to 5, we can obtain the patient's preference utility for the expert, the applicant's preference utility for the expert, and the expert's preference utility for the patient, and introduce the variable x ij Construct a multi-objective optimization model, x ij The value is 1 or 0. If xij If it is equal to 1, the patient and the expert are matched; otherwise, they are not. The mathematical formula group representing the multi-objective optimization model is as follows:

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025] x ij ={0, 1}, i = 1, 2,..., m; j = 1, 2,..., n; (Formula 26g)

[0026] Among them, Formulas 26a to 26d are the objectives of the multi-objective optimization model. Formulas 26a to 26c respectively satisfy the maximization of the preference utility of the patient, the expert, and the applying doctor. Formula 26d satisfies the minimization of the difference between the supply side and the demand side. Formula 26e indicates that the expert provides services within the service capacity. Formula 26f indicates that the patient will surely enjoy the medical services of the expert, that is, the patient must be matched with an expert; Z1 represents the patient preference utility value, Z2 represents the patient preference utility value, Z3 represents the patient preference utility value, Z4 represents the difference between the supply side and the demand side, m represents the maximum value of the number of patients, n represents the maximum value of the number of experts, P represents the patient, S represents the expert, R represents the applying doctor, P i represents the i-th patient, R i represents the applying doctor providing medical services for the i-th patient, i takes values from 1 to m, S j represents the j-th expert, j takes values from 1 to n, represents the influence intensity of the applying doctor R i on the patient P i , that is, the grey correlation coefficient, represents the preference utility of the applying doctor R i on the expert S j , represents the preference utility of the patient P i on the expert S j , represents the preference utility of the expert S j on the patient P i , y j represents the total number of services provided by the j-th expert within the service capacity;

[0027] The matching results for each matching entity are calculated according to the multi-objective optimization model.

[0028] Preferably, the one-to-many matching method means that one applying doctor can be in charge of multiple patients, and one patient has only one applying doctor; one expert can provide medical services for multiple patients, and one patient can only receive the services of one expert.

[0029] Preferably, when performing step 1, the attribute types of heterogeneous information include exact number type, interval number type, and language information type. The attribute values of the attribute types include expected values and evaluation values, and the expected values and evaluation values of the same attribute type are in the same information form;

[0030] The attribute values of the evaluation values of heterogeneous information are represented by exact numbers, interval numbers, or linguistic variables.

[0031] Preferably, when performing step 2, when calculating the distance measure of heterogeneous information, it is carried out in four different scenarios, which are respectively that the attribute values of the evaluation values are all exact numbers, the attribute values of the evaluation values are all interval numbers, the attribute values of the evaluation values are all exact linguistic evaluation information, and the attribute values of the evaluation values are all uncertain linguistic evaluation information.

[0032] Preferably, when performing step 3, the attribute weights for prospect value integration are determined through the evaluation information given by the matching entity. The smaller the difference in the attribute values under the same attribute type, the smaller the influence of the attribute type on the decision-making, and the smaller the attribute weight; conversely, the larger the attribute weight.

[0033] Preferably, when performing step 6, the multi-objective model is converted into a single-objective model through the objective function linear weighting method, and then the mathematical optimization software LINGO or Cplex is used for solution to obtain the optimal matching result.

[0034] Preferably, when performing step 1, the patient, the applying doctor, and the expert respectively submit their expected information about the matching object to the telemedicine information service platform; the applying doctor is the attending doctor of the patient, understands the patient's condition and situation, and the applying doctor evaluates the patient and submits the evaluation information to the telemedicine information service platform; the expert evaluation information is given by the telemedicine information service platform according to historical data or submitted by the expert to the telemedicine information service platform.

[0035] Preferably, the patient, the applying doctor, and the expert all submit the expected information or evaluation information through their respective local client servers, and the telemedicine information service platform is a cloud server platform.

[0036] A method for matching supply and demand in telemedicine considering doctor-patient relationship and intermediary intervention according to the present invention solves the technical problem of supply-demand matching in the context of telemedicine. Starting from the characteristics and requirements of the supply-demand matching decision-making problem in telemedicine, by analyzing the bounded rationality characteristics, psychological behaviors of matching subjects, and the interaction relationship between two-level demanders, a supply-demand matching decision-making method considering a two-level demand network is given. The present invention analyzes the supply-demand matching problem of two-level demands, which has obvious characteristics compared with the existing methods that only consider the supply and demand sides and the supply-demand matching problem considering a third party. It fully considers each participating subject and its preference in the matching decision-making process, is an extension of the bilateral matching theory in the complex background of telemedicine, makes up for the deficiencies of current research, and maximizes the preference utility of supply-demand subjects. Considering the evaluation preferences and habits of decision-makers in the real situation, the present invention simultaneously considers the certainty and ambiguity of decision-maker evaluation, integrates evaluation information of multiple data types to construct a matching decision model, and enhances the feasibility and rationality of decision-making compared with most existing studies that consider a single form of evaluation type. In the actual situation, the professionalism of medical knowledge makes patients more or less influenced by doctors in the medical decision-making process. The present invention uses the peer effect to explain this phenomenon and uses grey relational analysis to describe the mutual relationship and influence intensity, making the decision model closer to reality and thus guiding practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is the matching framework diagram of the telemedicine service of the present invention;

[0038] Figure 2 is the flowchart of the present invention. DETAILED IMPLEMENTATION MANNER

[0039] By Figure 1 - Figure 2 A method for matching supply and demand in telemedicine considering doctor-patient relationship and intermediary intervention shown, includes the following steps:

[0040] Step 1: Collect heterogeneous information provided by the supply and demand sides participating in the telemedicine service through the telemedicine information service platform; the supply and demand sides include multiple matching subjects; the types of the matching subjects include the demander being a patient or an applying doctor, the supplier being an expert, the applying doctor being the attending doctor of the patient, and both the applying doctor and the patient, and the expert and the patient adopt a one-to-many matching method for matching;

[0041] Such as Figure 1The telemedicine supply and demand matching process is shown, in which the matching between the applicant doctor and the patient, and between the expert and the patient is one-to-many. That is, one applicant doctor can be in charge of multiple patients, and one patient has only one in-charge doctor; one expert can provide medical services to multiple patients, and one patient can only receive services from one expert. Before matching, the patient has completed the first consultation in the hospital where he is located. Therefore, one expert can guide multiple applicant doctors, and one applicant doctor can also apply for consultation with multiple experts, that is, the expert and the applicant doctor are in a many-to-many relationship.

[0042] The telemedicine information service platform integrates relevant heterogeneous information, matches the supply and demand sides according to their preferences, and recommends the best matching pairs.

[0043] Heterogeneous information includes the expected information and evaluation information of each matching subject on its matching object. In the heterogeneous information, the mutual evaluation matrix of the matching subjects is constructed according to the evaluation information, and the expectation matrix of each matching subject is constructed according to the expected information. Figure 1 The diagram shows the telemedicine service matching framework, where P = {P i |i=1,2,...,m} is the patient set, P i represents the i-th patient; since there is a one-to-many relationship between the applicant doctor and the patient, for the convenience of representation, the applicant doctor set is defined as R = {R i |i=1,2,...,m},R i represents the doctor who provides the first consultation service for the i-th patient; S = {S j |j=1,2...,n} is the expert set, S j represents the jth expert.

[0044] definition is a set of attributes for evaluating patients, represents the g-th attribute of the patient, is the attribute set of the evaluation expert, represents the kth attribute of the expert. The expert’s preference weight for the patient’s attribute is represents the expert's weight preference for the g-th attribute of the patient, The patient's preference weight for the expert's attributes is represents the patient's weighted preference for the expert's k-th attribute, The preference weight of the applicant doctor for the expert attribute is represents the weight preference of the applicant doctor for the kth attribute of the expert, In actual decision-making problems, the evaluations of decision-makers often involve mixed multi-attribute decision-making problems that contain quantitative and qualitative information. Quantitative information includes exact numerical values, interval numbers, etc., and qualitative information includes linguistic evaluation information, etc. Therefore, this invention considers three types of attribute types: exact numbers, interval numbers, and linguistic information, and the expected value and evaluation value of the same attribute are in the same information form. Let C PK 、C PI and C PL respectively represent the patient attributes with exact numbers, interval numbers, and linguistic information as attribute values, and C PK ∪C PI ∪C PL =C P ;

[0045] Similarly, C SK ∪C SI ∪C SL =C S .

[0046] The decision-maker is everyone who provides evaluation information, and can be an expert, a primary care doctor, or a patient.

[0047] The expected information of the patient for the expert is indicating the expected value of the k-th attribute of the patient P i for the expert; the expected information of the applying doctor for the expert indicating the expected value of the k-th attribute of the applying doctor R i for the expert; the expected information of the expert for the patient indicating the expected value of the g-th attribute of the expert S j for the patient; the evaluation information of the actual situation of the patient indicating the actual evaluation value of the g-th attribute of the patient P i ; the actual evaluation information of the expert indicating the actual evaluation value of the k-th attribute of the expert S j . In addition, the duration of telemedicine services is random. To ensure the workload balance of experts, this invention balances the workload of each expert by the number of patients received, and defines y j as the maximum number of patients received by the expert S j .

[0048] In this embodiment, and are taken as examples to illustrate the construction process of the telemedicine supply-demand matching decision model.

[0049] Step 2: Calculate the distance measure for heterogeneous information with the expected values of each matching entity as the reference points to obtain the distance measure results, and use the gap between the actual evaluation value and the expected value in the heterogeneous information as the condition for measuring the preference utility of each matching entity;

[0050] According to prospect theory, decision-makers will select the optimal alternative based on the prospect values of all alternatives. The establishment and change of the reference point affect people's feelings of gains and losses, and thus affect decision-making. In the present invention, using the expected value as the reference point can accurately express the characteristics of the bounded rational behavior of decision-makers and well integrate various properties of prospect theory.

[0051] Let be the gap between the actual value of the k-th attribute of the expert and the expected value of patient P i for this attribute. When the actual value of the expert is higher than the patient's expectation, i.e., When Otherwise,

[0052] Scenario 1: That is, and are both exact numbers.

[0053] To eliminate the influence of dimension, the decision-making information needs to be normalized before calculating the distance between the actual value and the expected value. When the attribute value is an exact number, its normalization calculation formula is:

[0054]

[0055] Then the distance measure between the actual value and the expected value is expressed as:

[0056]

[0057] Scenario 2: That is, and are both interval numbers. Its normalization formula is as follows:

[0058]

[0059]

[0060] The magnitude of interval numbers is compared using the possibility method:

[0061]

[0062] If Then It means that the actual evaluation value of the expert attribute is higher than the patient's expected value, i.e., If Then It means that the patient's expected value is equal to the actual evaluation value of the expert, i.e., Conversely, the patient's expected value is higher than the actual evaluation value of the expert, i.e., Then its distance measure function is shown in Formula 6 below:

[0063]

[0064] Scenario 3: And And are both exact language evaluation information.

[0065] Define as the subscript value of the language evaluation value , as the subscript value of the language evaluation value . Then its normalization formula and distance function are shown in Formula 7 and Formula 8 respectively.

[0066]

[0067]

[0068] Scenario 4: And And are both uncertain language evaluation information.

[0069] In real decision-making, due to the fuzziness and uncertainty of thinking, decision-makers may show hesitation among several evaluation terms, and the hesitation degrees of each decision-maker are not the same, that is, l p ≠l s , where l p and l s are the numbers of elements in the decision-making sets of patients and experts respectively. When measuring the distance of hesitant fuzzy language, the supplementation method is usually adopted, that is, by some means, evaluation values are supplemented in the set with fewer numbers until l p =l s . However, this method is too subjective and easily causes distortion of decision-making information. Different extensions will obtain different distances, thus affecting the decision-making results. The present invention measures the distance between hesitant fuzzy language information based on the distance function: Definition 1: Define N S ={S δk |k = 1, 2,..., l} as a hesitant fuzzy language information set on the language term set S={S α |α=-τ,..., -1, 0, 1,..., τ}, τ is a positive integer and is the upper bound of the language term set. Then the score function and deviation function of H S are defined as where δ k is the subscript of the hesitant fuzzy language information set, δ k ∈S α .

[0070] Definition 2: Definition and are any two hesitant fuzzy linguistic information sets on S = {S α |α = -τ,..., -1, 0, 1,..., τ}, then and The distance between them is expressed as:

[0071]

[0072] where I(H S ) represents the subscript of the linguistic term S δk . For and If then If and then If and then If and then Based on this, and The distance measure between them is expressed as:

[0073]

[0074] Step 3: According to the distance measure results, establish the profit and loss matrix of each matching subject relative to the reference point, calculate the benefits and losses of each matching subject according to the profit and loss matrix, and establish the prospect decision matrix of each matching subject under the condition of bounded rationality by considering the different psychological states of the matching subject towards benefits and losses;

[0075] Represent the preference utility of the matching subject by integrating the prospect values in the prospect decision matrix;

[0076] Based on the distance measure of heterogeneous information in Step 2, the profit and loss matrix of the patient relative to the reference point can be established When the actual value of the expert attribute is higher than the patient's expectation, the patient benefits; otherwise, when the actual value cannot meet the patient's needs, losses occur. Considering the different psychological states of the patient towards benefits and losses, the prospect decision matrix under the condition of bounded rationality can be established where is the prospect value of the k-th attribute of the expert for the patient P i , and is expressed as:

[0077]

[0078] where represents the benefit, Denote loss; α, β are risk preference coefficients, satisfying 0 ≤ α, β ≤ 1. The larger the values of α and β, the more risk - prone the decision - maker is. λ is the loss - aversion coefficient, and λ > 1 indicates that the decision - maker is sensitive to loss risk. In prospect theory, the relevant parameter values can represent the general behavior preferences of decision - makers. α = β = 0.88 and λ = 2.25. After that, to understand patient P i For expert S j 's preference, the present invention represents the preference utility of the patient by integrating the patient's prospect value. The preference utility function of patient P i for expert S j is shown in Equation 12:

[0079]

[0080] where, represents the weight information of the k - th attribute of the patient for the expert. To overcome the subjectivity of weight - determination methods such as the analytic hierarchy process, the present invention determines the attribute weights according to the evaluation information given by the evaluation subject. The maximum deviation method points out that the smaller the difference in attribute values under the same attribute, the smaller the impact of the attribute on the decision - making, and the smaller the attribute weight; otherwise, the larger the attribute weight. Attribute weights play an important role in decision - making ranking and scheme selection.

[0081] In this embodiment, the attribute weights are obtained by solving the following optimization model:

[0082]

[0083] where, is the attribute weight before normalization, represents the difference in attribute values between any two schemes under the same attribute. The distance measure method in Step 2 is used to calculate the distance of heterogeneous information. Then, the above - mentioned optimization model is solved and normalized to obtain the following attribute weights:

[0084]

[0085] Step 4: Measure the psychological perception of each matching subject on the matching result through disappointment theory, so as to correct the preference utility of each matching subject;

[0086] The disappointment theory is used to describe the phenomenon that the behavior of individuals under risk conditions deviates from the results predicted by the expected utility theory, and it performs well in explaining and describing individual psychological perceptions. Its basic idea is that decision-makers will compare the expected results with their expectations. If the decision result is lower than the psychological expectation, the decision-maker will show disappointed psychological behavior; otherwise, they will show joy. When the magnitude of the difference is the same, disappointment has a greater impact on an individual's preference than joy, and this phenomenon is called disappointment aversion. Therefore, to fully consider the psychological behavior of the subject in the matching decision, the present invention introduces the disappointment theory to measure the psychological perception of the matching subject towards the matching result and modifies the preference utility calculation method.

[0087] An individual's actual utility is a modified utility, including: the traditional utility function and the disappointment-joy function. Assume there are n natural states s1, s2,..., s n , s i represents the i-th state, and p i represents the probability of s i occurring, where 0 ≤ p i ≤ 1, and Let x1, x2,..., x n respectively represent the results of the occurrence of s1, s2,..., s n states, then the modified utility of x i is expressed as:

[0088] u(x i ) = v(x i ) + D(v(x i ) - e[v(X)]) (Formula 15)

[0089] Among them, v(x i ) is the utility obtained by the individual from the result x i , e[v(X)] is the expected utility value, D(·) is a monotonically increasing disappointment-joy function that satisfies D(0) = 0. If v(x i ) - e[v(X)] > 0, then D(·) captures the individual's happiness value; otherwise, it captures the individual's disappointment value.

[0090] A single expected expectation may be ambiguous, and an individual's disappointment-joy value is related to all possible other results. If all the results are sorted in descending order of preference as x1 ≥ x2 ≥... ≥ x n , then the new modified utility function is:

[0091]

[0092] D(·) and E(·) are concave non - decreasing functions that capture disappointment and elation sensitivities respectively, with D(0) = 0 and E(0) = 0. D(·) can be expressed as:

[0093] D(x)=1 - θ Δx , Δx≥0 (Equation 17)

[0094] where Δx is the utility difference, 0 < θ < 1 is a parameter reflecting the concavity of the disappointment function, and the disappointment value decreases as θ increases. The value range of the parameter is [0.7, 0.9]. The disappointment value of the patient for the matching result can be obtained according to Equation 17. In other words, given the matching pair (P i , S j ), the preference utility of the patient is such that Then patient P i will be disappointed because of being matched with expert S j instead of being matched with expert S t because S t brings a higher utility to the patient. Let be the disappointment value obtained by P i being matched with S j instead of being matched with S t . can be expressed as:

[0095]

[0096] Before the matching result is formed, the probability of patient P i being matched with each expert is equal. Let Then the comprehensive disappointment value obtained by P i being matched with S j is:

[0097]

[0098] In addition, E(·) can be expressed as:

[0099]

[0100] where Δx is the utility difference, is a parameter reflecting the concavity of the elation function, and the elation degree decreases as increases. The value range of i is [0.7, 0.9]. γ is a disappointment aversion parameter, satisfying 0 < γ < 1, and the disappointment aversion degree increases as γ decreases. The elation value of the patient for the matching result can be obtained according to Equation 20. In other words, given the matching pair (P i , S j ), such that Then patient P i Will be due to expert S j Matching without expert S t I am happy to match with S j Matching enables patients to obtain higher preference utility. Indicates P i With S j Matches without S t The joy value obtained by matching, It can be expressed as:

[0101]

[0102] Similarly, before the matching result is formed, patient P i The likelihood of matching with each expert is equal.

[0103] make Then P i With S j The comprehensive joy value obtained by matching is:

[0104]

[0105] Based on the above analysis and combined with disappointment theory, we can get patient P i The modified preference utility function of is:

[0106]

[0107] Step 5: Based on grey correlation analysis, measure the influence of different doctors on patients’ decision-making, take patients’ expectations of experts as reference sequence, and improve the application of grey correlation function in heterogeneous information according to the distance between the expected information of doctors and the reference sequence;

[0108] In a telemedicine scenario, the patient and the applicant doctor are both on the demand side, and because the applicant doctor has rich clinical experience and medical knowledge, the patient's decision will be influenced by his or her attending physician to a certain extent. At the same time, due to different individual characteristics, different individuals have different degrees of influence on patients. Therefore, this embodiment introduces grey correlation analysis to measure the degree of influence of different applicant doctors on patients, because grey correlation analysis can reflect changes in sequence trends. This embodiment uses the patient's expectations of experts as a reference sequence, and uses the distance between the applicant doctor's expected information and the reference sequence to improve the application of the grey correlation function in heterogeneous information, as shown in Formula 24:

[0109]

[0110] in, Indicates application for doctor R i The expert has the same k-th attribute as the patient Pi The grey correlation coefficient, where ρ ∈ [0, 1] is the resolution coefficient, usually taken as 0.5. Then, for the applying doctor R i on the patient P i the degree of influence is expressed as:

[0111]

[0112] Step 6: According to the methods in Steps 1 to 5, obtain the preference utility of the patient for the experts, the preference utility of the applying doctor for the experts, and the preference utility of the experts for the patient. Introduce the variable x ij to construct a multi-objective optimization model, where x ij takes values of 1 or 0. If x ij = 1, then the patient is matched with the expert; otherwise, not. The mathematical formula group representing the multi-objective optimization model is as follows:

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] x ij = {0, 1}, i = 1, 2,..., m; j = 1, 2,..., n; (Formula 26g)

[0120] Among them, Formulas 26a to 26d are the objectives of the multi-objective optimization model. Formulas 26a to 26c respectively satisfy the maximization of the preference utilities of the patient, the expert, and the applying doctor. Formula 26d satisfies the minimization of the difference between the supply side and the demand side. Formula 26e indicates that the expert provides services within the service capacity. Formula 26f indicates that the patient will surely enjoy the medical services of the expert, that is, the patient must be matched with an expert; Z1 represents the patient preference utility value, Z2 represents the patient preference utility value, Z3 represents the patient preference utility value, Z4 represents the difference between the supply side and the demand side, m represents the maximum value of the number of patients, n represents the maximum value of the number of experts, P represents the patient, S represents the expert, R represents the applying doctor, P i represents the i-th patient, R i represents the applying doctor providing medical services for the i-th patient, where i takes values from 1 to m, S j represents the j-th expert, where j takes values from 1 to n, represents the applying doctor R iThe influence intensity on patient P i , i.e., the grey correlation coefficient, represents the preference utility of applying doctor R i for expert S j ; represents the preference utility of patient P i for expert S j ; represents the preference utility of expert S j for patient P i , y; j represents the total number of services provided by the j-th expert within the service capacity;

[0121] The matching results for each matching entity are calculated according to the multi-objective optimization model.

[0122] In this embodiment, according to the methods of steps 1 to 5, it can be obtained that:

[0123] The preference utility of patient P i for expert S j is where is the revised preference effect of the patient, is the revised preference utility of the applying doctor; the preference utility of applying doctor R i for expert S j is The preference utility of expert S j for patient P i is To solve the problem of remote medical supply-demand matching, the present invention introduces a 0-1 variable x ij to construct a multi-objective optimization model. If x ij = 1, patient P i is matched with expert S j , otherwise not.

[0124] The present invention can maximize the preference utility of the supply-demand entities. The remote medical supply-demand matching process involves patients, applying doctors, and experts. Among them, the applying doctor is both the demand side and the mediator of remote medical services and plays an important role in the development and operation of remote medicine. Therefore, the preferences of applying doctors should also be considered in remote medical supply-demand matching.

[0125] When performing step 1, the attribute types of heterogeneous information include exact number type, interval number type, and language information type. The attribute values of the attribute types include expected values and evaluation values, and the expected values and evaluation values of the same attribute type are in the same information form;

[0126] The attribute values of the evaluation values of heterogeneous information are represented by exact numbers, interval numbers, or linguistic variables.

[0127] When performing step 2 and calculating the distance measure of heterogeneous information, it is carried out in four different scenarios, which are respectively that the attribute values of the evaluation values are all exact numbers, the attribute values of the evaluation values are all interval numbers, the attribute values of the evaluation values are all exact language evaluation information, and the attribute values of the evaluation values are all uncertain language evaluation information.

[0128] When performing step 3, the attribute weights during the integration of prospect values are determined by matching the evaluation information given by the subject. The smaller the difference in attribute values under the same attribute type, the smaller the influence of the attribute type on the decision-making and the smaller the attribute weight; on the contrary, the larger the attribute weight.

[0129] When performing step 6, the multi-objective model is converted into a single-objective model by the linear weighted method of the objective function, and then the mathematical optimization software LINGO or Cplex is used for solution to obtain the optimal matching result.

[0130] When performing step 1, the patient, the applying doctor and the expert respectively submit their expected information about the matching object to the remote medical information service platform; the applying doctor is the attending doctor of the patient and understands the patient's condition and situation. The applying doctor evaluates the patient and submits the evaluation information to the remote medical information service platform; the expert evaluation information is given by the remote medical information service platform according to historical data or submitted by the expert to the remote medical information service platform.

[0131] The patient, the applying doctor and the expert all submit the expected information or evaluation information through their respective local client servers, and the remote medical information service platform is a cloud server platform.

[0132] The following is a specific example analysis of this embodiment:

[0133] Under a certain specialty at a certain stage, there are 6 patients P = {P1, P2,..., P6} under the leadership of their attending doctors R = {R1, R2,..., R6} seeking the help of experts on the remote medical service platform. There are 4 experts S = {S1, S2, S3, S4} on this platform who can provide remote medical services, and the maximum workload of each expert is 2. The attributes of the experts and the expectations of the patients and applying doctors for the experts mainly consider the expert reputation (C1 S ) and professional level treatment effect waiting time The attributes of the patients and the expectations of the experts for the patients mainly consider the tolerance (C1 P ) and the severity of the illness communication ability Among them, the attribute is a cost type, and the others are benefit type indicators, and the attribute C1 SThe evaluation information and expected information are represented by exact numbers (unit: percentage), The evaluation information and expected information are represented by interval numbers (unit: hours), C1 P and The evaluation information and expected information are represented by exact language terms, and The evaluation information and expected information are represented by uncertain language information, and the set of language terms used is S = {S0, S1, S2, S3, S4, S5, S6} = {very poor, poor, bad, medium, good, very good, excellent}. The evaluation information of the applying doctor on the patient and the expected information of the expert on the patient are shown in Table 1 Evaluation Information of Patients and Expected Information of Experts on Patients. The evaluation information of the expert given by the telemedicine service platform is shown in Table 2 Evaluation Information of Experts. The expected information of the patient and the applying doctor on the expert is shown in Table 3 Expected Information of Patients and Applying Doctors on Experts.

[0134]

[0135] Table 1

[0136]

[0137] Table 2

[0138]

[0139] Table 3

[0140] To solve the above problems, first, it is necessary to normalize and measure the distance of the evaluation information and expected information in various expressions under different attributes according to Formula 1 - Formula 10 to measure the positive and negative gaps between the evaluation information and the expected reference point. Secondly, according to the expected information given by the evaluation subject, the preference utility of each subject for the matching object is calculated by combining Formula 11 - Formula 14, where the preference weights of each subject for relevant attributes are obtained from Formula 13 - Formula 14, that is, the weight preferences of the expert for the 3 attributes of the patient are (0.328, 0.180, 0.492), the weight preference of the patient for the expert attributes is (0.302, 0.366, 0.078, 0.254), and the weight preference of the applying doctor for the expert attributes is (0.393, 0.214, 0.085, 0.308). Then, based on the disappointment theory, the psychological perception of the matching subject is measured to correct the preference utility function, taking γ = 0.5, and the final preference utility matrices of the patient, the applying doctor, and the expert are obtained, as shown in Table 4 and Table 5. Table 4 shows the corrected preference utility of the patient and the applying doctor, and Table 5 shows the corrected preference utility of the expert.

[0141] <![CDATA[S1]]> <![CDATA[S2]]> <![CDATA[S3]]> <![CDATA[S4]]> <![CDATA[S1]]> <![CDATA[S2]]> <![CDATA[S3]]> <![CDATA[S4]]> <![CDATA[P1]]> -0.043 0.090 0.508 -0.001 <![CDATA[R1]]> 0.426 0.287 0.613 -0.274 <![CDATA[P2]]> 0.141 -0.181 0.429 -0.177 <![CDATA[R2]]> 0.511 -0.129 0.674 0.082 <![CDATA[P3]]> 0.629 0.538 0.926 0.547 <![CDATA[R3]]> 0.092 -0.026 -0.210 -0.364 <![CDATA[P4]]> -0.444 -0.154 0.439 -0.199 <![CDATA[R4]]> 0.674 0.053 0.804 0.462 <![CDATA[P5]]> 0.465 0.122 0.732 0.259 <![CDATA[R5]]> 0.473 0.401 -0.062 -0.057 <![CDATA[P6]]> 0.274 0.145 0.572 0.024 <![CDATA[R6]]> 0.641 0.517 -0.105 0.413

[0142] Table 4

[0143] P1 P2 P3 P4 P5 P6 <![CDATA[S1]]> 0.273 0.738 0.387 -0.168 0.806 0.283 <![CDATA[S2]]> -0.239 0.332 0.016 -1.031 0.530 -0.138 <![CDATA[S3]]> 0.054 0.490 0.282 -0.423 0.717 0.078 <![CDATA[S4]]> 0.172 0.587 0.357 -0.180 0.746 0.253

[0144] Table 5

[0145] In addition, according to Formula 24 - Formula 25, the influence intensity of the applying doctor on the patient is ρ i R =(0.767, 0.745, 0.551, 0.606, 0.591, 0.676). Then, a multi - objective optimization model is constructed using Formula Group 26 and converted into a single - objective optimization model. Considering the fairness of each objective, the coefficients of each objective are taken to be equal, and the following optimal solution is obtained by solving through LINGO17.0: x 11 =1, x 23 =1, x 32 =1, x 44 =1, x 51 =1, x 64 =1, and the remaining x ij =0, that is, patients P1, P5 are matched with expert S1, patient P2 is matched with expert S3, patient P3 is matched with expert S2, and patients P4, P6 are matched with expert S4.

[0146] A remote - medical supply - demand matching method considering doctor - patient relationship and intermediary intervention according to the present invention solves the technical problem of supply - demand matching in the context of remote medical treatment. Starting from the characteristics and requirements of the supply - demand matching decision problem in the remote medical service, by analyzing the bounded - rationality characteristics, psychological behaviors of the matching subjects, and the interaction relationship between two - level demanders, a supply - demand matching decision method considering a two - level demand network is given. The present invention analyzes the supply - demand matching problem of two - level demands, and has obvious characteristics compared with the existing methods that only consider the supply and demand sides and the supply - demand matching problem considering the third party. It fully considers each participating subject and its preference in the matching decision - making process, is an extension of the bilateral matching theory in the complex background of remote medical treatment, makes up for the deficiencies of the current research, and maximizes the preference utility of the supply - demand subjects; considering the evaluation preferences and habits of decision - makers in the real situation, the present invention simultaneously considers the certainty and ambiguity of decision - maker evaluation, integrates evaluation information of multiple data types to construct a matching decision model, and enhances the feasibility and rationality of the decision compared with most existing studies that consider a single - form evaluation type; in the actual situation, the professionalism of medical knowledge makes patients more or less influenced by doctors in the medical treatment decision - making process. The present invention uses the peer effect to explain this phenomenon and uses grey relational analysis to describe the mutual relationship and influence intensity, making the decision model closer to reality and then guiding practice.

Claims

1. A method for matching the supply and demand of telemedicine considering the doctor-patient relationship and intermediary intervention, characterized in that: The steps include: Step 1: Collect heterogeneous information provided by the supply and demand parties participating in the telemedicine service through the telemedicine information service platform; the supply and demand parties include multiple matching subjects; the types of matching subjects include demand parties and supply parties, the demand parties are patients or applicant doctors, the supply parties are experts, the applicant doctors are the patients' attending doctors, and the applicant doctors and patients, experts and patients are matched using a one-to-many matching method; Heterogeneous information includes the expectation information and evaluation information of each matching subject on its matching object; in the heterogeneous information, the mutual evaluation matrix of the matching subjects is constructed according to the evaluation information, and the expectation matrix of each matching subject is constructed according to the expectation information; Step 2: Take the expected value of each matching subject as the reference point to calculate the distance measurement of heterogeneous information, and obtain the distance measurement result. The gap between the actual evaluation value and the expected value in the heterogeneous information is used as the condition to measure the preference utility of each matching subject. Let be the difference between the actual value of the k-th attribute of the expert and the expected value of patient P i for that attribute; when the expert's actual value is higher than the patient's expectation, i.e., When Otherwise, Scenario 1: That is and are both exact numbers; In order to eliminate the impact of dimension, the decision information needs to be normalized before calculating the distance between the actual value and the expected value; when the attribute value is an exact number, the normalized calculation formula is: Then the distance measure between the actual value and the expected value is expressed as: Scenario 2: That is and are both interval numbers, and their normalization formula is as follows: The size of the interval numbers is compared using the possibility method: If then it means that the actual evaluation value of the expert attribute is higher than the patient's expectation value, that is If then it means that the patient's expectation value is equal to the actual evaluation value of the expert, that is On the contrary, the patient's expectation value is higher than the actual evaluation value of the expert, that is then its distance measure function is as follows: Scenario 3: and and are both exact language evaluation information; Definition is the subscript value of the language evaluation value , and is the subscript value of the language evaluation value , then its normalization formula and distance function are shown as follows; Scenario 4: and and are both uncertain linguistic evaluation information; The distance between hesitant fuzzy language information is measured based on the distance function: Definition 1: Define H S ={S δk |k = 1, 2, ..., l} is a hesitant fuzzy linguistic information set on the linguistic term set S = {S α |α = -τ, ..., -1, 0, 1, ..., τ}, where τ is a positive integer and is the upper bound of the linguistic term set. Then the score function and deviation function of H S are defined as follows where δ k is the subscript of the hesitant fuzzy linguistic information set, δ k ∈ S α ; Definition 2: Definition and are any two hesitant fuzzy linguistic information sets on S = {S α |α = -τ,..., -1, 0, 1,..., τ}, then and The distance between them is expressed as: where, I(H S ) represents the subscript of the language term S δk ; for and if then if and then if and then if and then Based on this, and the distance measure between them is expressed as: Step 3: Based on the distance measurement results, establish the profit and loss matrix of each matching subject relative to the reference point, calculate the gains and losses of each matching subject based on the profit and loss matrix, and consider the different psychological states of the matching subject towards gains and losses, and establish the prospect decision matrix of each matching subject under the condition of limited rationality; The preferred utility of the matching subject is represented by integrating the prospect values in the prospect decision matrix; Step 4: Use the disappointment theory to measure each matching subject's psychological perception of the matching results, so as to modify the preference utility of each matching subject; Step 5: Based on grey correlation analysis, measure the influence of different doctors on patients’ decision making, take patients’ expectations of experts as reference sequence, and improve the application of grey correlation function in heterogeneous information according to the distance between the expected information of doctors and the reference sequence; Step 6: Obtain the preference utility of the patient for the expert, the preference utility of the applicant doctor for the expert, and the preference utility of the expert for the patient according to the methods in Steps 1 to 5, and introduce the variable x ij Construct a multi-objective optimization model, x ij takes a value of 1 or 0. If x ij = 1, the patient is matched with the expert; otherwise, they are not matched. The mathematical formula group representing the multi-objective optimization model is as follows: x ij = {0, 1}, i = 1, 2, ..., m; j = 1, 2, ..., n; (Equation 26g) Among them, formulas 26a to 26d are the objectives of the multi-objective optimization model. Formulas 26a to 26c respectively satisfy the maximization of the preference utilities of patients, experts, and applying doctors. Formula 26d satisfies the minimization of the difference degree between the supply side and the demand side. Formula 26e indicates that experts provide services within their service capabilities, and formula 26f indicates that patients will surely enjoy the medical services of experts, that is, patients must be matched with an expert; Z1 represents the patient preference utility value, Z2 represents the patient preference utility value, Z3 represents the patient preference utility value, Z4 represents the difference degree between the supply side and the demand side, m represents the maximum value of the number of patients, n represents the maximum value of the number of experts, P represents patients, S represents experts, R represents applying doctors, and Pi i represents the i-th patient. R i denotes the applying doctor who provides medical services for the \(i\)-th patient, where \(i\) ranges from 1 to \(m\), \(S\) j denotes the \(j\)-th expert, where \(j\) ranges from 1 to \(n\), denotes the applying doctor \(R\) i 's impact intensity on patient \(P\) i , i.e., the grey correlation coefficient, denotes the preference utility of the applying doctor \(R\) i for expert \(S\) j ; denotes the preference utility of patient \(P\) i for expert \(S\) j ; denotes the preference utility of expert \(S\) j for patient \(P\) i , \(y\) j denotes the total number of services provided by the \(j\)-th expert within the service capacity; The matching results for each matching subject are calculated according to the multi-objective optimization model.

2. The remote medical supply-demand matching method considering doctor-patient relationship and intermediary intervention as described in claim 1, characterized in that: The one-to-many matching method means that one applicant doctor can be in charge of multiple patients, and one patient can only have one applicant doctor; one expert can provide medical services to multiple patients, and one patient can only receive services from one expert.

3. The remote medical supply-demand matching method considering the doctor-patient relationship and intermediary intervention as described in claim 1, characterized in that: When executing step 1, the attribute types of the heterogeneous information include an exact number type, an interval number type, and a language information type, the attribute values of the attribute types include an expected value and an evaluation value, and the expected value and the evaluation value of the same attribute type are in the same information form; The attribute values of the evaluation values of heterogeneous information are expressed by exact numbers, interval numbers or linguistic variables.

4. The remote medical supply-demand matching method considering the doctor-patient relationship and intermediary intervention according to claim 3, wherein: When executing step 2, when calculating the distance measurement for heterogeneous information, it is performed in four different scenarios. The four different scenarios are: the attribute values of the evaluation values are all exact numbers, the attribute values of the evaluation values are all interval numbers, the attribute values of the evaluation values are all exact language evaluation information, and the attribute values of the evaluation values are all uncertain language evaluation information.

5. The remote medical supply-demand matching method considering doctor-patient relationship and intermediary intervention according to claim 4, wherein: When performing step 3, the attribute weights during the integration of foreground values are determined by matching the evaluation information given by the main body. The smaller the difference in attribute values under the same attribute type, the smaller the influence of the attribute type on the decision-making, and the smaller the attribute weight; On the contrary, the greater the attribute weight.

6. The remote medical supply-demand matching method considering the doctor-patient relationship and intermediary intervention according to claim 1, wherein: When performing step 6, the multi-objective model is converted into a single-objective model by the linear weighted method of the objective function, and then the mathematical optimization software LINGO or Cplex is used for solution to obtain the optimal matching result.

7. A method for matching the supply and demand of telemedicine considering the doctor-patient relationship and intermediary intervention according to claim 1, characterized in that: When performing step 1, the patient, the applying doctor, and the expert respectively submit their expected information about the matching object to the telemedicine information service platform; the applying doctor is the patient's attending doctor who understands the patient's condition and situation. The applying doctor evaluates the patient and submits the evaluation information to the telemedicine information service platform; the expert evaluation information is given by the telemedicine information service platform according to historical data or submitted by the expert to the telemedicine information service platform.

8. The telemedicine supply-demand matching method considering the doctor-patient relationship and intermediary intervention according to claim 7, characterized in that: The patient, the applying doctor, and the expert all submit the expected information or evaluation information through their respective local client servers, and the telemedicine information service platform is a cloud server platform.

Images