Medical service total element efficiency evaluation improvement method for reflecting case complexity and realizing Pareto optimum

Abstract

The invention provides a medical service total element efficiency evaluation improvement method capable of reflecting case complexity and realizing Pareto optimization. The method comprises the following steps: step 1, collecting data; step 2, diagnosing related groups; step 3, risk adjustment; step 4, DEA modeling is carried out; step 5, data input and operation; and 6, obtaining an evaluation result. Compared with most of other performance evaluation methods based on case grouping (such as DRG, DIP and the like), the technical scheme of the invention only considers the output of the medical service, such as the discharge number of people, the medical cost index, the hospitalization duration index and the low-risk hospitalization death rate, and does not consider the investment of doctors, nurses, beds and the like of the medical service. According to the invention, not only one medical service output is considered, but also multiple outputs are considered; the method not only considers the resource consumption (such as case weight) relative to the average level, but also considers various investments, and is a comprehensive multi-input and multi-output total factor productivity evaluation method.

Description

Technical Field

[0001] The present invention relates to the fields of medical service management, performance evaluation, total factor productivity, and productivity measurement technology, and in particular to an improved method for evaluating the total factor efficiency of medical services that reflects case complexity and achieves Pareto optimality. Background Art

[0002] (I) The Diagnosis Related Group (DRG) performance evaluation method is commonly used in medical services. Although the case weight of this method measures the resource consumption of diseases relative to the average level, it does not take into account the input factors of medical service providers (such as medical staff, medical facilities, etc.) and cannot measure the overall service efficiency of the evaluated decision-making unit (DMU).

[0003] (II) The performance evaluation method that can comprehensively consider input and output factors is the data envelopment analysis (DEA) method, which can evaluate the comprehensive efficiency under multiple inputs and multiple outputs. However, when this method is applied to the field of medical services, it only evaluates the performance of medical services based on the number of discharges, and does not consider the differences in medical service performance caused by the heterogeneity of case complexity. For example, the Department of Stomatology performs 5 dental surgeries a day, and the Department of Cardiothoracic Surgery performs 1 heart bypass surgery a day. The case complexity of the two is different, but without considering the case complexity, the former is more efficient than the latter in terms of quantity alone, which is obviously unreasonable.

[0004] (III) Under the guidance of the current idea of ​​"taking people's health as the center", the performance evaluation of medical inpatient services should not only consider expected outputs such as the number of discharged patients, but also consider non-expected outputs such as hospitalization costs, length of stay, and low-risk inpatient mortality, so as to more comprehensively evaluate the performance of medical services. To cure the same type of disease, some medical groups spend an average of 10,000 yuan, while some medical groups spend an average of 12,000 yuan. Obviously, the former has higher performance. Similarly, to cure the same type of disease, some medical groups require patients to stay in the hospital for an average of 5 days, while some medical groups require patients to stay in the hospital for an average of 7 days. Obviously, the former has higher performance. Furthermore, the inpatient mortality rate is also an important indicator for measuring medical services. The lower this indicator is, the better. Ignoring this indicator is obviously one-sided. In addition, some diseases have a higher risk of inpatient mortality, such as rare malignant cancers in the elderly; some diseases have a lower risk of inpatient mortality, such as the inpatient mortality rate of young pregnant women. It is biased if the difference in the risk of death from different diseases is not considered when evaluating inpatient mortality. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide an improved method for evaluating the total factor efficiency of medical services that reflects the complexity of cases and achieves Pareto optimality. It not only considers one medical service output, but also multiple outputs; it not only considers resource consumption relative to the average level, but also multiple inputs. It is a comprehensive total factor productivity evaluation method with multiple inputs and multiple outputs.

[0006] To achieve the above-mentioned purpose, the present invention adopts the following technical scheme: an improved method for evaluating the total factor efficiency of medical services that reflects case complexity and achieves Pareto optimality, comprising the following steps: Step 1: collecting data; Step 2: diagnosis-related grouping; Step 3: risk adjustment; Step 4: DEA modeling; Step 5: data input and calculation; Step 6: obtaining evaluation results.

[0007] In a preferred embodiment, the step 1 specifically includes: collecting inpatient service evaluation index data from the evaluated decision-making unit (DMU); in terms of input, selecting the number of doctors and nurses from the dimension of human input, and selecting the number of beds from the dimension of asset investment; in terms of output, selecting discharges from the dimension of capacity, selecting the average length of stay and average hospitalization expenses from the dimension of efficiency, and selecting the number of inpatient deaths IpD from the perspective of quality.

[0008] In a preferred embodiment, step 2 specifically includes: for a group of DMUs being evaluated, the same diagnosis-related grouping criteria are adopted; the cases are divided into disease groups according to "similar clinical significance and similar resource usage", and different weights are assigned to each group; the higher the severity and complexity of the disease, the longer the hospitalization time, and the more expensive the treatment cost, the higher the weight, and vice versa; the length of hospitalization, hospitalization cost, and mortality rate of different cases are homogeneous within the group, and the weights are adjusted between groups, thereby solving the problem of case complexity heterogeneity of medical services.

[0009] In a preferred embodiment, the step 3 specifically includes: dividing all cases into Q diseases by diagnosis-related grouping, Q ≥ 1; assuming that the set of all diseases is Ω, and each disease is a subset Ω of Ω g , g=1,2,…,Q, the average hospitalization cost of all diseases g is The average length of hospital stay is

[0010] Assume that a certain unit DMU is j ,j=1,2,…,J cases are divided into M diseases, 1≤M≤Q, let the set of M diseases be Φ, Each disease is a subset of Φ g , It is also the seed set of all cases. gA subset of Get the average hospitalization cost of each disease type g in the evaluated unit Average length of hospital stay The ratios of medical expenses and hospitalization duration of each disease type g, g = 1, 2, ..., M of the evaluated unit to the medical expenses and hospitalization duration of the corresponding disease type of all cases are:

[0011] Medical expense ratio

[0012] Hospital stay ratio

[0013] The superscript c represents medical expenses, and the subscript g represents disease type g; The superscript t represents the length of hospital stay, and the subscript g represents the disease type g;

[0014] The ratio of medical expenses for the disease type g of the evaluated unit to the total medical expenses The weighted average is used to obtain the medical cost index, which is:

[0015] Medical Cost Index

[0016] Where m g It represents the number of cases of each disease type g in the evaluated unit;

[0017] The ratio of the hospitalization time of the disease type g (g = 1, 2, ..., M) of the evaluated unit to the total hospitalization time The weighted average is used to obtain the hospital stay index, which is:

[0018] Hospital stay index

[0019] In step 2, all disease groups g, g = 1, 2, ..., Q, are assigned a weight w according to the severity, complexity, length of hospital stay, and cost of treatment. g ; For each disease type g (g = 1, 2, ..., M) of the evaluated unit, its case mix index CMI is

[0020]

[0021] Where p g Represents the number of cases of all diseases g, g = 1, 2, ..., Q;

[0022] CMI means case complexity, CMI = 1 means the complexity is at the average level, CMI = 1.5 means the complexity is 1.5 times the average level; CMI is used as a risk adjustment parameter to adjust the number of discharges, hospital stay length index LSI and medical expenses index MEI. The formula is as follows:

[0023] aDischarges=CMI×Discharges (5)

[0024] aMEI=MEI / CMI (6)

[0025] aLSI=LSI / CMI (7)

[0026] The larger the adjusted number of discharges aDischarges is, the more complicated cases are discharged; the larger the adjusted medical expense index aMEI is, the more the cost of treatment for cases of the same complexity is higher than the average level; the larger the adjusted length of stay index aLSI is, the more the time spent on treatment for cases of the same complexity is higher than the average level;

[0027] Define the mortality risk level of each disease; in step 2, summarize all hospitalized deaths by disease type, obtain the total number of hospitalized deaths for each disease type g, and divide it by the total number of hospitalized people to obtain the hospital mortality rate D g According to the statistical probability characteristics, the number of hospitalized deaths and hospital mortality rate satisfy the Poisson distribution, so the natural logarithm of the hospital mortality rate for each disease is taken, that is, ln(D g ) ; then calculate the mean of the logarithm of the inpatient mortality rate of all diseases, that is and standard deviation σ g , and then according to D g , and σ g Define the death risk level; a death risk level of "0" indicates that there is no death risk in these disease groups, also known as zero risk; a death risk level of "1" indicates that the logarithm of the mortality rate in these disease groups is less than its mean minus 1 standard deviation, also known as the low-risk group LRG; a death risk level of "2" indicates that the logarithm of the mortality rate in these disease groups is between its mean and the mean minus 1 standard deviation, also known as the medium-low risk group MLRG; a death risk level of "3" indicates that the logarithm of the mortality rate is between its mean and the mean plus 1 standard deviation, also known as the medium-high risk group MHRG; a death risk level of "4" indicates that the logarithm of the mortality rate is greater than its mean plus 1 standard deviation, also known as the high-risk group HRG; through the classification of death risk levels, it can be seen that the probability of death in the disease groups of levels "0" and "1" is extremely low; if death occurs at these two levels, the possibility of improper surgery and / or care is higher.

[0028] In a preferred embodiment, step 4 specifically includes: first, constructing a production possibility set (PPS) of inputs and outputs; assuming that there are J comparable DMUs, each DMU has I inputs and R outputs; for the jth DMU, ​​the quantity of its i-th input is x ij , the quantity of the rth output is recorded as y rj , DMU j The input and output can be recorded as: X = (x 1j ,x 2j ,…,x ij ,…,…,x Ij ) T and Y=(y 1j ,y 2j ,…,y rj ,…,y Rj ) T ; Under the standard assumptions of convexity and free disposability of inputs and outputs, the production possibility set formed by the set of J DMUs is as follows:

[0029]

[0030] The production possibility set of DMU k with respect to super-efficiency is expressed as That is, the production possibility set consisting of all DMUs except DMU k is expressed as follows:

[0031]

[0032] in The superscript s indicates super efficiency, and the subscript k indicates DMU k;

[0033] Secondly, under the VRS assumption, the model saves and bounded output surplus Determine DMU k and frontier surface The minimum distance, the model of the first stage is expressed as:

[0034]

[0035] in and The ranges of input and output are defined as:

[0036]

[0037] In the second phase, the savings will be Added to the input of the super-efficient DMU k and subtracted from the output of the super-efficient DMU k Push DMU k to production frontiers other than its own;

[0038] To generate a Pareto efficient projection, model (11) determines the additional possible bounded input surplus and bounded output deficiency The model of the second stage is expressed as:

[0039]

[0040] Finally, the efficiency scores of the above models (10) and (11) are combined to calculate the final comprehensive efficiency score according to the following formula (12):

[0041]

[0042] The DMUs are then ranked according to their efficiency scores.

[0043] In a preferred embodiment, step 5 specifically includes: inputting the input data of all decision-making units (DMUs), such as the number of doctors, the number of nurses, and the number of beds, the expected output data, such as the adjusted number of discharges aDischarges, and the unexpected output data, such as the adjusted medical expense index aMEI, the adjusted length of stay index aLSI, and the number of inpatient deaths in the low-risk group IDLRG, into the DEA model, treating the unexpected output as a negative number; DEA is a relative efficiency based on multiple outputs divided by multiple inputs on the production frontier, and ratio values ​​should be used with caution in the DEA model.

[0044] In a preferred embodiment, step 6 specifically includes: the input data is grouped according to diagnosis and risk adjusted, and the model (10) is first run to obtain the efficiency score δ of the DMU in the first stage. 1 , investment savings Output Surplus Run model (11) again to obtain the efficiency score of each DMU in the second phase 2 , Overinvestment Insufficient output Finally, run the model (12) to get the final score Then according to Ranking of DMU;

[0045] The production frontier obtained after running models (10) and (11) is Pareto optimal; the relaxation generated by running the first stage model (10) Shows the advantages and amount of input savings and output surplus of the evaluated DMU in the evaluation group; runs the slack generated by the second stage model (11) Point out the excess input and insufficient output of the evaluated DMU, ​​and provide the direction and amount of performance improvement.

[0046] The present invention also provides a computer-readable storage medium, comprising: a computer program is stored on the computer-readable storage medium; when the computer program is run by a processor, the method for improving the evaluation of the total factor efficiency of medical services that reflects case complexity and achieves Pareto optimality is executed.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] (I) Unlike most other performance evaluation methods based on case grouping (such as DRG, DIP, etc.), this application only considers the output of medical services, such as the number of discharges, medical cost index, length of stay index, and low-risk hospital mortality rate, and lacks consideration of inputs such as doctors, nurses, and beds for medical services. This application considers not only one medical service output, but also multiple outputs; not only considers resource consumption relative to the average level, but also considers multiple inputs. It is a comprehensive multi-input and multi-output total factor productivity evaluation method.

[0049] (II) Unlike other methods that use DEA to evaluate medical service performance, this application only considers the number of discharges and does not consider the differences in medical service performance caused by the heterogeneity of case complexity. This application not only considers expected outputs, including the number of discharges that reflect case complexity; it also considers non-expected outputs, including the hospitalization duration index, hospitalization cost index, and the number of hospitalization deaths in the low-risk disease group.

[0050] (III) The total factor productivity evaluation method of this application can achieve Pareto Optimality. First, it helps decision makers recognize their own advantages and give quantitative values; second, it helps decision makers discover their own disadvantages, find the direction of improvement, and indicate the amount of improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the steps of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0052] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0053] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0054] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0055] The technical solution of this application is as follows Figure 1 As shown, the following steps are included: (i) data collection; (ii) DRG grouping; (iii) risk adjustment; (iv) DEA modeling; (v) data input and calculation; and (vi) obtaining evaluation results.

[0056] 1. Data Collection

[0057] Inpatient service evaluation index data were collected from the evaluated decision-making units (DMUs). In terms of input, the number of doctors and nurses was selected from the dimension of human input, and the number of beds was selected from the dimension of asset input. In terms of output, the number of discharges was selected from the dimension of capacity, the average length of stay and average hospitalization expenses were selected from the dimension of efficiency, and the number of inpatient deaths (IpD) was selected from the perspective of quality.

[0058] 2. Diagnosis-related groups

[0059] With the development and derivation of diagnosis-related groups (DRGs), the World Health Organization used case-based groups (CBGs) in 2020. [1] As a broader general term for DRG. The CBG standards used in my country mainly include CN-DRG jointly formulated by the Medical Administration and Management Bureau of the National Health Commission and the Information Center of the Beijing Municipal Health Commission, C-DRG formulated by the Health Development Research Center of the National Health Commission, CHS-DRG formulated by the National Healthcare Security Administration, and the Diagnosis-Intervention Packet (DIP) formulated by the National Healthcare Security Administration. These are all based on the International Classification of Diseases (ICD) codes, including the main diagnosis and secondary diagnosis, treatment type, patient age and gender, surgery, the presence of comorbidities and complications (CC), discharge status and the surgery performed. Different standards differ in specific case grouping, disease coding, disease weight, medical insurance subsidy method and subsidy amount calculation, but their principles are the same.

[0060] For a group of DMUs being evaluated, the same diagnosis-related grouping criteria should be used. Cases are divided into disease groups based on "similar clinical significance and similar resource usage", and each group is assigned a different weight. The higher the severity and complexity of the disease, the longer the hospital stay, and the more expensive the treatment cost, the higher the weight, and vice versa. The length of hospital stay, hospitalization costs, and mortality rates of different cases are homogeneous within the group, and the weights are adjusted between groups, which solves the problem of case complexity heterogeneity in medical services.

[0061] (III) Risk Adjustment

[0062] By diagnosis-related grouping, assume that all cases can be divided into Q diseases (Q ≥ 1). Let the set of all diseases be Ω, and each disease be a subset of Ω g=1,2,…,Q),

[0063] The average hospitalization cost of all diseases g (g = 1, 2, ..., Q) can be obtained Average length of hospital stay Assume that a certain unit DMU is j (j=1,2,…,J) cases can be divided into M diseases (1≤M≤Q). Let the set of M diseases be Each disease is a subset of Φ g Φ g It is also the seed set of all cases. g A subset of The average hospitalization cost of each disease type g (g = 1, 2, ..., M) of the evaluated unit can be obtained Average length of hospital stay The ratios of medical expenses and hospitalization duration of each disease type g (g = 1, 2, ..., M) in the evaluated unit to the medical expenses and hospitalization duration of the corresponding disease type of all cases are:

[0064] Medical expense ratio

[0065] Hospital stay ratio

[0066] The superscript c represents medical expenses, and the subscript g represents disease type g; The superscript t represents the length of hospital stay, and the subscript g represents the disease type g.

[0067] The ratio of medical expenses for all diseases g (g = 1, 2, ..., M) in the evaluated unit to the total medical expenses Taking weighted average, we get:

[0068] Medical Cost Index

[0069] The ratio of the hospitalization time of all diseases g (g = 1, 2, ..., M) in the evaluated unit to the hospitalization time of all cases Taking weighted average, we get

[0070] Hospital stay index

[0071] Where m g Represents the number of cases of each disease type g (g=1,2,…,M) in the evaluated unit.

[0072] In step “(ii) diagnosis-related grouping”, all disease groups g (g = 1, 2, …, Q) are assigned a weight W according to the severity, complexity, length of hospital stay, and cost of treatment. g For each disease type g (g = 1, 2, ..., M) of the evaluated unit, its case mixed index (CMI) is

[0073]

[0074] Where n g Represents the number of cases of all diseases g (g=1,2,…,Q).

[0075] CMI means case complexity. A CMI of 1 means the complexity is at the average level, and a CMI of 1.5 means the complexity is 1.5 times the average level. CMI is used as a risk adjustment parameter to adjust the number of discharges, the length of stay index (LSI) and the medical expenses index (MEI). The formula is as follows:

[0076] aDischarges=CMI×Discharges (5)

[0077] aMEI=MEI / CMI (6)

[0078] aLSI=LSI / CMI (7)

[0079] The larger the adjusted discharge number (aDischarges, adjusted Discharges), the more complex cases are discharged; the larger the adjusted medical expense index (aMEI, adjusted MEI), the higher the cost of treatment for cases of the same complexity is than the average level; the larger the adjusted length of stay index (aLSI, adjusted LSI), the higher the time spent on treatment for cases of the same complexity is than the average level.

[0080] In addition, due to the heterogeneity of case complexity, it is difficult to distinguish which case death is caused by the disease itself and its severity, and which case death is caused by improper surgery and / or care. Therefore, we define the mortality risk level of the disease. In step "(ii) Diagnosis-related grouping", all inpatient deaths are summarized according to each disease type, and the number of inpatient deaths of all diseases g (g = 1, 2, ..., Q) is obtained, and this is divided by the total number of inpatients to obtain the inpatient mortality rate D g According to the statistical probability characteristics, the number of hospitalized deaths and hospital mortality rate satisfy the Poisson distribution. Therefore, the natural logarithm of the hospital mortality rate for each disease is taken, that is, ln(D g ). Then calculate the mean of the logarithm of the inpatient mortality rate of all diseases, that is, and standard deviation σ g , and then according to D g , and σ g The mortality risk levels are defined as shown in Table 1 below:

[0081] Table 1 Definition of mortality risk levels

[0082]

[0083] The death risk level of "0" means that there is no death risk in these disease groups, also known as zero risk; "1" means that the logarithm of the mortality rate of these disease groups is less than its mean minus 1 standard deviation, also known as low risk group (LRG); "2" means that the logarithm of the mortality rate of these disease groups is between its mean and mean minus 1 standard deviation, also known as medium-to-low risk group (MLRG); "3" means that the logarithm of the mortality rate is between its mean and mean plus 1 standard deviation, also known as medium-to-high risk group (MHRG); "4" means that the logarithm of the mortality rate is greater than its mean plus 1 standard deviation, also known as high risk group (HRG). Through the classification of death risk levels, it can be seen that the probability of death in disease groups with levels "0" and "1" is extremely low; if death occurs at these two levels, then the possibility of improper surgery and / or care is higher, so it is more scientific and objective to evaluate the quality of medical services.

[0084] (IV) DEA Modeling

[0085] In order to achieve a full ranking of inpatient service performance, this application proposes a super-efficiency data envelopment analysis (DEA) model that can achieve Pareto optimality. This model comprehensively evaluates inpatient service performance, taking into account not only expected outputs (such as discharges) but also undesirable outputs (such as costs, time, and hospital deaths). In this model, undesirable outputs are treated as negative values.

[0086] In many cases, the size of the decision-making units (DMUs) of medical services varies greatly. For example, some DMUs have dozens of doctors, while others have only a single digit. Therefore, the variable return to scale (VRS) assumption is adopted. In addition, the VRS assumption is also translation invariant, which avoids distortion of the evaluation results when dealing with negative values. DEA requires careful selection of inputs and outputs, because efficiency is likely to be affected by the definition of outputs and the number of inputs and outputs included. In the DEA model, inputs and undesired outputs are usually "poor" types of performance measures, or low-quality indicators; while expected outputs are usually "better" types of performance measures, or high-quality indicators. For zero values ​​in output data, if you want to reflect the actual efficiency of a DMU not producing a certain output (for example, the number of inpatient deaths in a DMU is 0), a non-radial model should be used. Therefore, we choose the non-radial slacks-based measure (SBM) model as the basic DEA model for performance evaluation. On the one hand, the SBM model can eliminate non-radial slack. When the efficiency score is used as the only indicator to evaluate the performance of DMU, ​​the remaining non-radial slack in the radial model may mislead the decision. In addition, the SBM model can be used to construct a non-directional model, assuming that the decision maker can control both inputs and outputs. This scenario is more suitable for medical and health practice.

[0087] Assume there are J comparable DMUs, each with I inputs and R outputs. For the jth DMU, ​​the quantity of its i-th input is x ij , the quantity of the rth output is recorded as y rj , DMU j The input and output can be recorded as: X = (x 1j ,x 2j ,…,x ij ,…,…,x Ij ) T and Y=(y 1j ,y 2j ,…,y rj ,…,y Rj ) T Under the standard assumptions of convexity and free disposability of inputs and outputs, the production possibility set formed by the set of J DMUs is as follows:

[0088]

[0089] The production possibility set of DMU k with respect to super-efficiency is expressed as That is, the production possibility set consisting of all DMUs except DMU k is expressed as follows:

[0090]

[0091] in The superscript S represents super efficiency, and the subscript k represents DMU k.

[0092] Secondly, under the VRS assumption, the model saves and bounded output surplus Determine DMU k and frontier surface The minimum distance, the model of the first stage is expressed as:

[0093]

[0094] in and The ranges of input and output are defined as:

[0095]

[0096] In the second phase, the savings will be Added to the input of the super-efficient DMU k and subtracted from the output of the super-efficient DMU k Push DMU k to production frontiers other than its own.

[0097] To generate the Pareto efficient projection, model (11) identifies additional possible bounded input surpluses and bounded output deficiency The model of the second stage is expressed as:

[0098]

[0099] Finally, the efficiency scores of the above models (10) and (11) are combined to calculate the final comprehensive efficiency score according to the following formula (12):

[0100]

[0101] The DMUs are then ranked according to their efficiency scores.

[0102] (V) Model calculation

[0103] The input data of comparable decision-making units (DMUs, such as between medical groups or between departments or between hospitals) such as the number of doctors, nurses, and beds, the expected output data such as adjusted discharges (aDischarges), and the undesirable output data such as adjusted medical expense index (aMEI), adjusted length of stay index (aLSI), and inpatient deaths of low-risk group (Inpatient Deaths of Low-Risk Group, IDLRG), with undesirable outputs treated as negative numbers, are input into the DEA model. DEA is a relative efficiency based on multiple outputs divided by multiple inputs on the production frontier. Ratio values ​​should be used with caution in the DEA model. Therefore, we use low-risk inpatient deaths instead of low-risk inpatient mortality rates in the DEA model.

[0104] (VI) Obtaining evaluation results

[0105] The input data, after diagnosis-related grouping and risk adjustment, first runs model (10) to obtain the efficiency score δ of the DMU in the first stage 1 , investment savings Output Surplus Run model (11) again to obtain the efficiency score δ of each DMU in the second stage 2 , Overinvestment Insufficient output Finally, run the model (12) to get the final score Then according to Ranking of DMU.

[0106] The production frontier obtained after running models (10) and (11) is Pareto optimal. The slack generated by running the first stage model (10) is The advantages and quantities of the input savings and output surplus of the evaluated DMU in the evaluation group are shown, which is helpful for decision makers to recognize and quantify their own advantages. The slack generated by running the second stage model (11) It points out the excess input and insufficient output of the evaluated DMU, ​​and provides the direction and amount of performance improvement.

Claims

1. An improved method for evaluating the total factor efficiency of medical services that reflects case complexity and achieves Pareto optimality, characterized in that: The method includes the following steps: Step 1: data collection; Step 2: diagnosis-related grouping; Step 3: risk adjustment; Step 4: DEA modeling; Step 5: data input and calculation; Step 6: obtaining evaluation results.

2. According to claim 1, a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality is characterized in that: The step 1 specifically includes: collecting inpatient service evaluation indicator data from the evaluated decision-making unit (DMU); in terms of input, selecting the number of doctors and nurses from the dimension of human input, and selecting the number of beds from the dimension of asset investment; in terms of output, selecting the number of discharges (Discharges) from the dimension of production capacity, selecting the average length of stay and average hospitalization expenses from the dimension of efficiency, and selecting the number of inpatient deaths (IpD) from the perspective of quality.

3. According to claim 1, a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality is characterized in that: The step 2 specifically includes: for a group of DMUs being evaluated, the same diagnosis-related grouping criteria are adopted; the cases are divided into disease groups according to "similar clinical significance and similar resource usage", and different weights are assigned to each group; the higher the severity and complexity of the disease, the longer the hospitalization time, and the more expensive the treatment cost, the higher the weight, and vice versa; the hospitalization time, hospitalization cost, and mortality rate of different cases are homogeneous within the group, and the weights are adjusted between groups, thus solving the problem of case complexity heterogeneity of medical services.

4. According to claim 1, a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality is characterized in that: The step 3 specifically includes: dividing all cases into Q diseases through diagnosis-related grouping, where Q≥1; assuming that the set of all diseases is Ω, and each disease is a subset Ω of Ω g , The average hospitalization cost for all diseases g is The average length of hospital stay is Assume that a certain unit DMU is j ,j=1,2,…,J cases are divided into M diseases, 1≤M≤Q, let the set of M diseases be Φ, Each disease is a subset of Φ g , Φ g It is also the seed set of all cases. g A subset of Get the average hospitalization cost of each disease type g in the evaluated unit Average length of hospital stay The ratios of medical expenses and hospitalization duration of each disease type g, g = 1, 2, ..., M of the evaluated unit to the medical expenses and hospitalization duration of the corresponding disease type of all cases are: Medical expense ratio Hospital stay ratio The superscript c represents medical expenses, and the subscript g represents disease type g; The superscript t represents the length of hospital stay, and the subscript g represents the disease type g; The ratio of medical expenses for the disease type g of the evaluated unit to the total medical expenses The weighted average is used to obtain the medical cost index, which is: Medical Cost Index Where m g It represents the number of cases of each disease type g in the evaluated unit; The ratio of the hospitalization time of the disease type g (g = 1, 2, ..., M) of the evaluated unit to the total hospitalization time The weighted average is used to obtain the hospital stay index, which is: Hospital stay index In step 2, all disease groups g, g = 1, 2, ..., Q, are assigned a weight w according to the severity, complexity, length of hospital stay, and cost of treatment. g ; For each disease type g (g = 1, 2, ..., M) of the evaluated unit, its case mix index CMI is Where p g Represents the number of cases of all diseases g, g = 1, 2, ..., Q; CMI means case complexity, CMI = 1 means the complexity is at the average level, CMI = 1.5 means the complexity is 1.5 times the average level; CMI is used as a risk adjustment parameter to adjust the number of discharges, hospital stay length index LSI and medical expenses index MEI. The formula is as follows: aDischarges=CMI×Discharges (5) aMEI=MEI / CMI (6) aLSI=LSI / CMI (7) The larger the adjusted number of discharges aDischarges is, the more complicated cases are discharged; the larger the adjusted medical expense index aMEI is, the more the cost of treatment for cases of the same complexity is higher than the average level; the larger the adjusted length of stay index aLSI is, the more the time spent on treatment for cases of the same complexity is higher than the average level; Define the mortality risk level of each disease; in step 2, summarize all hospitalized deaths by disease type, obtain the total number of hospitalized deaths for each disease type g, and divide it by the total number of hospitalized people to obtain the hospital mortality rate D g According to the statistical probability characteristics, the number of hospital deaths and hospital mortality rate satisfy the Poisson distribution, so the natural logarithm of the hospital mortality rate for each disease is taken, that is, ln(D g ) ; then calculate the mean of the logarithm of the inpatient mortality rate of all diseases, that is and standard deviation σ g , and then according to D g , and σ g The death risk level was defined as "0", which means there is no death risk in these disease groups, also known as zero risk; "1" means the logarithm of the mortality rate in these disease groups is less than the mean minus 1 standard deviation, also known as the low-risk group LRG; "2" means the logarithm of the mortality rate in these disease groups is between the mean and the mean minus 1 standard deviation, also known as the medium-low risk group MLRG; "3" means the logarithm of the mortality rate is between the mean and the mean plus 1 standard deviation, also known as the medium-high risk group MHRG; "4" means the logarithm of the mortality rate is greater than the mean plus 1 standard deviation, also known as the high-risk group HRG; through the classification of death risk levels, it can be seen that the probability of death in the disease groups of levels "0" and "1" is extremely low; if death occurs at these two levels, the possibility of improper surgery and / or care is higher.

5. According to claim 1, a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality is characterized in that: The step 4 specifically includes: first, constructing a production possibility set (PPS) of inputs and outputs; assuming that there are J comparable DMUs, each of which has I inputs and R outputs; for the jth DMU, ​​the quantity of its i-th input is x ij , the quantity of the rth output is recorded as y rj , DMU j The input and output can be recorded as: X = (x 1j ,x 2j ,…,x ij ,…,…,x Ij ) T and Y=(y 1j ,y 2j ,…,y rj ,…,y Rj ) T ; Under the standard assumptions of convexity and free disposability of inputs and outputs, the production possibility set formed by the set of J DMUs is as follows: The production possibility set of DMU k with respect to super-efficiency is expressed as That is, the production possibility set consisting of all DMUs except DMU k is expressed as follows: in The superscript s indicates super efficiency, and the subscript k indicates DMU k; Secondly, under the VRS assumption, the model saves and bounded output surplus Determine DMU k and frontier surface The minimum distance, the model of the first stage is expressed as: in and The ranges of input and output are defined as: In the second phase, the savings will be Added to the input of the super-efficient DMU k and subtracted from the output of the super-efficient DMU k Push DMU k to production frontiers other than its own; To generate a Pareto efficient projection, model (11) determines the additional possible bounded input surplus and bounded output deficiency The model of the second stage is expressed as: Finally, the efficiency scores of the above models (10) and (11) are combined to calculate the final comprehensive efficiency score according to the following formula (12): The evaluated units DMU are then ranked according to their efficiency scores.

6. According to claim 1, a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality is characterized in that: The step 5 specifically includes: inputting the input data of comparable decision-making units (DMUs) of the same nature, such as the number of doctors, the number of nurses, and the number of beds, expected output data, such as adjusted discharges (aDischarges), and undesirable output data, such as adjusted medical expense index (aMEI), adjusted length of stay index (aLSI), and inpatient deaths in the low-risk group (IDLRG), treating undesirable output as a negative number, into the DEA model; DEA is a relative efficiency based on multiple outputs divided by multiple inputs on the production frontier, and ratio values ​​should be used with caution in the DEA model.

7. According to claim 1, a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality is characterized in that: The step 6 specifically includes: inputting data, after diagnosis-related grouping and risk adjustment, first running the model (10) to obtain the efficiency score δ of the DMU in the first stage 1 , investment savings Output Surplus Run model (11) again to obtain the efficiency score δ of each DMU in the second stage 2 , Overinvestment Insufficient output Finally, run the model (12) to get the final score Then according to Ranking of DMU; The production frontier obtained after running models (10) and (11) is Pareto optimal; the relaxation generated by running the first stage model (10) Shows the advantages and amount of input savings and output surplus of the evaluated DMU in the evaluation group; runs the slack generated by the second stage model (11) Point out the excess input and insufficient output of the evaluated DMU, ​​and provide the direction and amount of performance improvement.

8. A computer-readable storage medium, characterized in that: include: The computer-readable storage medium stores a computer program; when the computer program is executed by a processor, it executes a method for improving the total factor efficiency evaluation of medical services that reflects case complexity and achieves Pareto optimality as described in any one of claims 1 to 7.

Images