Hospital comprehensive monitoring analysis method and platform based on artificial intelligence

Abstract

The invention belongs to the technical field of medical resource intelligent management, and relates to a hospital comprehensive monitoring analysis method and platform based on artificial intelligence, and the method comprises the following steps: generating a medicine management basic data set; generating a primary prediction signal; adjusting a prediction weight according to the conflict level and outputting a reliable demand prediction instruction; a quantitative evaluation value reflecting the medicine turnover efficiency, the immediate pressure and the cost loss is calculated, and when the quantitative evaluation value exceeds a preset safety alert threshold value, an inventory abnormity alarm signal is generated; outputting an allocation strategy instruction containing the medicine flow direction, the allocation quantity and the time efficiency; and comparing a deviation value between the reliable demand prediction instruction and the allocation strategy instruction, and when the deviation value continuously exceeds a preset deviation threshold value, triggering an update prompt of the drug safety use limit rule base. The problem that demand prediction and clinical rule verification are separated in a traditional mode is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent management of medical resources, and relates to a hospital comprehensive monitoring and analysis method and platform based on artificial intelligence. Background Art

[0002] Currently, the management of the hospital drug supply chain faces severe dual challenges of fluctuating clinical demands and inventory risks. The uncertainty of medical behaviors, sudden environmental factors, and complex drug safety rules lead to frequent imbalances between drug supply and demand, manifested as coexistence of shortages of critical drugs and expiration and scrapping of high-value drugs. It is difficult to respond in real time to sudden changes in the demands of clinical departments, and the fragmentation of inventory information among pharmacy nodes further amplifies the risk of resource misallocation. The traditional prediction and scheduling mode dominated by manual experience has significant lag in dealing with multi-dimensional interference factors.

[0003] The existing technologies mainly rely on static rule libraries and independent optimization modules for local improvements. Typical solutions include time series prediction models based on historical consumption, inventory alarm systems that set fixed safety inventory thresholds, and ex-post control processes for manually reviewing drug contraindications. Some systems attempt to introduce basic logistics cost algorithms to generate transfer suggestions, but the decision-making process is disconnected from the actual clinical demands. These solutions usually adopt a phased serial processing mechanism, and data transfer among all links relies on batch import and export operations.

[0004] Based on the above problems, the traditional method has a disconnection between demand prediction and clinical rule verification, and cannot dynamically correct the prediction deviation caused by sudden changes in diagnosis and treatment behaviors, resulting in allocation conflicts of high-risk drugs. Summary of the Invention

[0005] In a first aspect, the present invention provides a hospital comprehensive monitoring and analysis method based on artificial intelligence, adopting the following technical solutions: A hospital comprehensive monitoring and analysis method based on artificial intelligence, comprising the following steps: S1. Obtain operation records, inventory status, medical behaviors, and external environment indicators involved in the hospital drug circulation, perform structured cleaning according to drug types, department units, and patient group types, and generate a basic drug management data set; S2. Based on the basic drug management data set, identify the characteristics of clinical diagnosis and treatment activities, the rules of inventory increase and decrease in the pharmacy, and environmental interference factors, calculate the expected consumption of various drugs by each department unit within a specified future period, and generate a primary prediction signal; S3. When a specific drug demand mutation is detected in the primary prediction signal, call the drug safe use limitation rule library for clinical rationality verification, adjust the prediction weight according to the conflict level, and output a reliable demand prediction instruction; S4. By continuously monitoring the batch inventory status of each pharmacy node, calculate the quantitative evaluation values reflecting drug turnover efficiency, approaching expiration pressure, and cost loss. When the quantitative evaluation values exceed the preset safety warning threshold, generate an inventory anomaly warning signal. S5. Receive reliable demand prediction instructions and inventory anomaly warning signals, dynamically generate a cross-warehouse transfer plan according to the preset cost priority and service guarantee requirements, and output transfer strategy instructions including drug flow direction, allocation quantity, and timeliness. S6. Obtain the actual drug consumption progress and the completion record of the scheduling task, compare the deviation values between the reliable demand prediction instructions and the transfer strategy instructions. When the deviation values continuously exceed the preset deviation threshold, trigger an update prompt for the drug safe use limitation rule library.

[0006] A further solution of the present invention is to generate a basic drug management data set, including the following steps: Collect the original data stream through a preset periodic scraping mechanism. The original data stream covers operation records, inventory status, medical behaviors, and external environment indicators involved in the hospital drug circulation, and perform a classification and completion operation on the missing fields. Uniformly convert the time stamp into the format of year, month, day, and hour, and map the description of the administration method in free text to a standard operation code. The converted original data stream is classified and reorganized according to three types of indexes: drug type, department unit, and patient group type. The data under the same type of index is arranged in chronological order to generate the columnar-stored basic drug management data set.

[0007] A further solution of the present invention is that the converted original data stream is classified and reorganized according to three types of indexes: drug type, department unit, and patient group type, including the following steps: The drug type dimension points to the unique drug identification identifier established based on the pharmacological classification code and dosage form specification, and the unique drug identification identifier is predefined by the hospital pharmacy department. The department unit dimension includes the outpatient pharmacy window number and the complete campus and building path information. The patient group type dimension is configured as a set of labels divided according to disease characteristics, and the label generation logic is defined as automatically marked if the patient's current main diagnosis code belongs to the cardiovascular disease classification catalog or the age is within the pediatric admission range.

[0008] A further solution of the present invention is to generate a primary prediction signal, including the following steps: Extract the clinical diagnosis and treatment activity characteristics, pharmacy inventory growth and decline rules, and environmental interference factors from the basic drug management data set, and convert them into a set of numerical impact factors. Adopt basic trend extrapolation to calculate the benchmark prediction quantity of each drug in the department unit. Superimpose the set of numerical impact factors on the baseline prediction according to the preset dynamic weight allocation principle to generate a primary prediction signal including an array of future multi-day consumption amounts; The primary prediction signal is embodied as a two-dimensional table data structure. The row index is the combined key of the drug code and the department unit path, the column index is the date sequence of the next seven days, and the cell value represents the expected consumption amount on that day.

[0009] A further solution of the present invention is to call the drug safety use limitation rule library for clinical rationality verification, including the following steps: When the predicted consumption amount for a single day exceeds the preset fluctuation threshold compared with the historical average value, it is determined as a specific drug demand mutation event; Associate the patient group label in the drug management basic data set with the list of contraindicated patient groups, the corresponding table of substitute drug relationships, and the medical insurance usage control threshold in the limitation rule library for conflict detection; According to the levels of severe conflict points, moderate conflict points or mild conflict points, selectively adjust the weight coefficients of clinical activity factors, inventory growth and decline factors, and environmental interference factors; The limitation rule library consists of an electronic rule list maintained by the hospital pharmacy department, including a set of disease classification codes corresponding to drug contraindications, records of pharmacological equivalent substitute drug mapping relationships, and a numerical table of the maximum daily payment quantity for each medical insurance category of drugs.

[0010] A further solution of the present invention is to generate an inventory exception warning signal, including the following steps: Calculate the drug turnover efficiency index, approaching expiration pressure index, and cost loss index; Synthesize a comprehensive risk score according to the preset weight coefficients, and the weight coefficients are determined based on the attribution analysis of historical risk events of the core objectives of drug management; When the comprehensive risk score exceeds the preset safety warning threshold associated with the pharmacy type, an inventory exception warning signal in JSON format is generated with an additional risk level label. The head of the inventory exception warning signal message contains the drug code and the pharmacy path code to which it belongs, the risk label is marked in the middle, and the description field of the main risk source is added at the end.

[0011] A further solution of the present invention is to dynamically generate a cross-library transfer plan according to the preset cost priority and service guarantee requirements, including the following steps: Perform a cross-library retrieval operation on the drugs with inventory exceptions, obtain the total available inventory in the whole hospital and match the safety inventory baseline value; For high-risk approaching expiration drugs, start the inventory rebalancing priority strategy, and generate a point-to-point direct transfer plan or a drug centralized redistribution process according to the upper limit of the inventory that can be released from the source pharmacy; The total score of the scheduling plan is calculated by the lowest total score method, and the total score of the scheduling plan is dynamically synthesized by the transfer quantity, the in-transit loss coefficient of drugs, the cumulative penalty item of transfer frequency, and the value factor of unit distance distribution cost.

[0012] A further solution of the present invention is to generate a point-to-point direct transfer plan or a drug centralized redistribution process according to the upper limit of the releasable inventory of the source pharmacy, including the following steps: The point-to-point direct transfer plan is applicable to the single-source and single-target scenario, and the transfer quantity takes the safety inventory gap value of the target node but does not exceed the upper limit of the releasable inventory of the source node; The drug centralized redistribution process is used for multi-source or multi-target scenarios, and the coordinated central pharmacy acts as a transit node to accept the re-secondary distribution of the inventory of each high-risk source to the gap nodes; The setting of the upper limit of the releasable inventory is based on the remaining quantity after deducting the safety inventory threshold from the total real-time inventory of drugs at the source node, and is superimposed with the priority release of near-expiry batches and the dynamic calibration of ensuring the supply of the basic needs of this node in the next week.

[0013] A further solution of the present invention is to trigger an update prompt for the drug safety use limitation rule library when the deviation value continuously exceeds the preset deviation threshold, including the following steps: Calculate three types of deviation indicators: the daily average consumption deviation rate, the dispensing difference value, and the change range of turnover efficiency; When any deviation indicator continuously exceeds the corresponding preset deviation threshold for N days, trace back the logical relationship between the entries in the limitation rule library and the execution process data; If a rule is detected to be missing or a parameter is outdated, generate a structured revision suggestion including a revision type code and an impact factor adjustment value; The deviation threshold is finally determined based on the deviation distribution characteristics between prediction and actual in historical operation data, superimposed with the drug criticality classification and management granularity requirements, and through simulating and verifying the sensitivity of the threshold to business risks.

[0014] In a second aspect, the present invention provides a hospital comprehensive monitoring and analysis platform based on artificial intelligence, adopting the following technical solutions: A hospital comprehensive monitoring and analysis platform based on artificial intelligence, including the following modules: A multi-source medical data collection module, which is used to obtain operation records, inventory status, medical behaviors, and external environment indicators involved in the drug circulation in the hospital, perform structured cleaning according to drug types, department units, and patient group types, and generate a basic drug management data set; A drug demand prediction module, which is used to identify clinical diagnosis and treatment activity characteristics, pharmacy inventory growth and decline laws, and environmental interference factors based on the basic drug management data set, calculate the expected consumption of various drugs in each department unit within a specified future period, and generate a primary prediction signal; A clinical rule verification module, which is used to call the drug safe use limitation rule library for clinical rationality verification when a specific drug demand mutation is detected in the primary prediction signal, adjust the prediction weight according to the conflict level, and output a reliable demand prediction instruction; An inventory risk scanning module, which continuously monitors the batch inventory status of each pharmacy node, calculates a quantitative evaluation value reflecting drug turnover efficiency, approaching expiration pressure, and cost loss, and generates an inventory exception warning signal when the quantitative evaluation value exceeds the preset safety warning threshold; A multi-objective scheduling decision-making module, which is used to receive the reliable demand prediction instruction and the inventory exception warning signal, dynamically generate a cross-library transfer plan according to the preset cost priority and service guarantee requirements, and output a transfer strategy instruction including drug flow direction, allocation quantity, and timeliness; An execution feedback analysis module, which obtains the actual drug consumption progress and the completion record of the scheduling task, compares the deviation value between the reliable demand prediction instruction and the transfer strategy instruction, and triggers an update prompt for the drug safe use limitation rule library when the deviation value continuously exceeds the preset deviation threshold.

[0015] In summary, the present invention includes the following beneficial technical effects: 1. By embedding the drug safe use limitation rule library into the prediction correction link, it automatically triggers contraindication matching and medical insurance rule verification when a demand mutation occurs, blocking the unreasonable allocation of high-risk drugs from the source. The mechanism of dynamically adjusting the prediction weight by the system effectively adapts to the changes in complex medical scenarios, avoids medication conflicts caused by sudden changes in medical treatment behaviors, and especially ensures the medication safety of special patient groups. The collaborative effect of the clinical rule verification module and multi-source data collection enables the drug scheduling decision to always comply with treatment specifications and regulatory requirements; 2. The execution feedback analysis module continuously tracks the deviation pattern between the prediction and the actual situation, and automatically generates a revision suggestion when it detects the failure of the rule library or new risk characteristics. By correlating multiple indicators such as the daily average consumption deviation rate, allocation difference value, and change range of turnover efficiency, the system establishes a dynamic calibration mechanism for the rule library parameters. This ability enables the hospital pharmacy management system to have the characteristics of continuous evolution, especially adapting to the changes in business rules brought about by the update of medical treatment technologies and the adjustment of medical insurance policies; 3. Through the dual-channel drive of the reliable demand prediction instruction and the inventory exception warning signal, the resource allocation is optimized on the premise of meeting clinical needs. The preset logic of service guarantee requirements and cost priorities ensures that emergency drugs are continuously supplied, and the total logistics cost is reduced by using the centralized redistribution process of drugs. This multi-objective coordination mechanism based on the characteristics of medical scenarios reduces the ineffective capital occupation while improving the drug accessibility, forming a Pareto improvement in the utilization of medical resources. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. The accompanying drawings are used to provide a further understanding of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0017] Figure 1 The flowchart in the embodiments of the present application is disclosed.

[0018] Figure 2 The structural schematic diagram in the embodiments of the present application is disclosed. Detailed implementation manners

[0019] In order to make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0020] The following combines the attached Figure 1 - Figure 2 Make a preferred and detailed description of the present invention.

[0021] Referring to the attached Figure 1 , the present invention proposes a hospital comprehensive monitoring and analysis method based on artificial intelligence, including the following steps: S1. Obtain the operation records, inventory status, medical behaviors, and external environment indicators involved in the drug circulation in the hospital, perform structured cleaning according to drug types, department units, and patient group types, and generate a basic drug management data set; S2. Based on the basic drug management data set, identify the characteristics of clinical diagnosis and treatment activities, the growth and decline rules of pharmacy inventory, and environmental interference factors, calculate the expected consumption of various drugs in each department unit within a specified future period, and generate a primary prediction signal; S3. When a specific drug demand mutation is detected in the primary prediction signal, call the preset drug safety use limit rule library for clinical rationality verification, adjust the prediction weight according to the conflict level, and output a reliable demand prediction instruction; S4. By continuously monitoring the batch inventory status of each pharmacy node, calculate the quantitative evaluation values reflecting drug turnover efficiency, approaching expiration pressure, and cost loss. When the quantitative evaluation value exceeds the preset safety warning threshold, generate an inventory exception warning signal; S5. Receive the reliable demand prediction instruction and the inventory exception warning signal, dynamically generate a cross-library transfer plan according to the preset cost priority and service guarantee requirements, and output a transfer strategy instruction including drug flow direction, allocation quantity, and timeliness; S6. Obtain the actual drug consumption progress and scheduling task completion records, compare the deviation values between the reliable demand forecast instructions and the allocation strategy instructions, and trigger an update prompt for the drug safety use limitation rule library when the deviation value continuously exceeds the preset deviation threshold.

[0022] In one embodiment of the present invention, step S1 includes the following steps: Data interfaces are deployed based on the actual operating environment of the medical institution, connecting various business systems. A pre-set periodic capture mechanism continuously captures raw data streams, covering operational records, inventory status, medical behavior, and external environmental indicators related to the hospital's drug flow. Operational records are designed as electronic evidence of the actual movement of medications, including scanned code records for drug shipments, infusion preparation execution times, and ward return confirmations. Inventory status is obtained by accessing real-time inventory interfaces from the drug warehouse and pharmacy terminals, including current available inventory quantities, unit prices for high-value consumables, and expiration dates for each drug batch. Medical behavior is derived from interactive data collected from doctor workstations and nurse operation terminals, including the time and content of electronic prescriptions, fields in test report review conclusions, and notes from doctors when switching medications. External environmental indicators represent public information accessed through the hospital's data center, including temperature and humidity monitoring values released by the local meteorological bureau and infectious disease risks reported on government health platforms.

[0023] The original data stream is structured and cleaned according to the three dimensions of drug type, department unit, and patient group type: the timestamp is uniformly adjusted to the year-month-day-time format, and the drug type refers to the unique drug identification identifier established based on the pharmacological classification code and dosage form specifications. The unique drug identification identifier is pre-defined by the hospital pharmacy department. The department unit dimension includes the outpatient pharmacy window number, complete campus and building path information. The patient group type dimension is configured as a set of labels divided by disease characteristics, and the label generation logic is defined as automatically marking if the patient's current main diagnosis code belongs to the cardiovascular disease classification catalog or the age is within the pediatric admission range. Structured cleaning refers to a specific process of performing classification and completion on the original data stream, including deleting duplicate entries that are completely identical, automatically filling in missing entries with system null value symbols, converting department abbreviations into full name dictionary items, and mapping free text descriptions of medication methods to standard operation codes.

[0024] The original converted data stream is classified and reorganized according to three types of indexes: drug types, department units, and patient group types. The data under the same index is arranged in chronological order. For records with missing patient group labels, the diagnostic codes of their source prescriptions are automatically associated. If the diagnostic codes belong to specific chronic disease or infectious disease classifications, they are marked as corresponding patient group entries. Finally, a basic drug management data set is generated, which is stored in columns according to the index columns, eliminates field ambiguities, and covers the complete transfer scenarios. The basic drug management data set is a columnar storage table that supports time range search. Each row record contains three mandatory fields: drug types, department units, and patient group types, as well as other affiliated information fields.

[0025] Exemplarily, the inpatient pharmacy business system performs an automatic scraping process every day at dawn. First, it collects the drug dispensing operation flow generated the previous day. For the drug record of "metformin" with a specification of 100mg 10 tablets / box, it replaces the drug name with the institutional standard code "YA003892". When it is found that the numerical value field of a certain drug dosage is empty, it automatically retrieves the average usage value of the past three days in its affiliated department, "Oncology Ward 3", as the filling.

[0026] The inpatient number of the prescription patient is associated with the electronic medical record system. If the patient is diagnosed with E11.9 (type 2 diabetes), the label "Chronic Metabolic Disease Patient Group" is added. After the data is aggregated according to the three index keys of "YA003892", "East Area, 9th Floor, Building B of the Inpatient Building", and "Chronic Metabolic Disease Patient Group", it is written into the central storage area to form the basic management entry of this drug on this date. The basic management entry of this drug on this date and other non-warning records together constitute the basic drug management data set of the day, and the basic drug management data set serves as the input basis for the entire process.

[0027] In one embodiment of the present invention, step S2 includes the following steps: Based on the basic drug management data set, identify the characteristics of clinical diagnosis and treatment activities, the growth and decline rules of pharmacy inventory, and environmental interference factors, calculate the expected consumption of various drugs in each department unit within a specified future period, and generate a primary prediction signal.

[0028] Specifically, the generated basic drug management data set is used as the only input source, and the historical data records of the three dimensions are split according to a preset time window.

[0029] According to the characteristics of clinical diagnosis and treatment activities, extract the total prescription quantity curves corresponding to the labels of various patient groups in each department unit, identify the abnormal bands with continuous increase or sudden drop among them, and set the characteristics of clinical diagnosis and treatment activities as the change rules of prescription behaviors associated with the labels of patient group types. Traverse the prescription operation records in the basic drug management dataset, calculate the increase or decrease amplitude of the prescription quantity of a specific patient group in the same department unit for three consecutive days. If the absolute value of the increase or decrease amplitude is greater than the increase or decrease amplitude threshold, it is recorded as an abnormal band. The increase or decrease amplitude threshold is set based on the statistical distribution of the increase or decrease amplitude of the prescription quantity of a specific patient group in the same department unit in the historical prescription data, by calculating the average increase or decrease amplitude plus twice the standard deviation (or adjusting the multiple according to business requirements).

[0030] According to the growth and decline rules of pharmacy inventory, count the change slope of the daily closing stock of each drug at the pharmacy node, excluding non-consumable changes such as inventory counts and adjustments. The growth and decline rules of pharmacy inventory are realized as a quantitative expression of the actual drug consumption rate. Specifically, select the difference in the daily closing stock of each pharmacy node in the basic drug management dataset, and calculate the sliding average of the consumption quantity change rate with a seven-day cycle after excluding the operation records of drug returns or losses.

[0031] Regarding environmental interference factors, operationally anchor the external environmental indicator fields in the basic drug management dataset, and mark the events where the temperature and humidity continuously exceed the warning threshold for two days or more as one valid interference record, and count the existence of infectious disease risks as one valid interference record.

[0032] The warning threshold is set based on the legal temperature and humidity range in the drug storage specification, combined with the technical parameters of the pharmacy environmental monitoring equipment and historical safe operation data, and determined by the upper limit values of the drug storage temperature and humidity specified by industry standards.

[0033] Convert the three types of recognition results into a set of numerical impact factors. The set of numerical impact factors is the numerical conversion form of the three types of recognition results. The clinical activity factor is the percentage value of the peak-to-valley difference in prescription quantity within the abnormal band accounting for the monthly average quantity. The inventory growth and decline factor is the integer representation value obtained by multiplying the consumption quantity change rate by the base number 100. The environmental interference factor directly uses the total number of valid interference records. Obtain the benchmark prediction quantity by performing basic trend extrapolation on the consumption sequence of each drug in each department. Basic trend extrapolation means taking the consumption quantity sequence of the drug in the department in the most recent 30 days and calculating the daily average consumption quantity as the benchmark prediction quantity. Then, multiply the three types of impact factors by their respective assigned weight coefficients and add them to the benchmark prediction quantity one by one.

[0034] The weight allocation principle adopts the principle of combining basic static allocation with dynamic condition triggering: the initial weights are pre-allocated with basic ratios according to the influence degree of each factor on the drug consumption prediction. When specific factor combinations (such as clinical activities and environmental interference factors) are detected to appear abnormally at the same time, the weight proportion of key factors is automatically increased according to the preset rules. For example, the initial configuration is that the clinical activity factor is 0.5, the inventory growth and decline factor is 0.3, and the environmental interference factor is 0.2. When the fluctuation value of the clinical activity factor > 30% and the environmental interference factor ≥ 3 are satisfied at the same time, the weight of the clinical activity factor is automatically increased by 25 percentage points to 0.75, and the remaining 0.25 weight is re-allocated to the inventory growth and decline factor and the environmental interference factor as 0.15 and 0.1. The initial weights are determined according to the influence degree of each factor on the prediction error through historical data regression analysis, and the thresholds in the dynamic adjustment mechanism (such as 30% fluctuation value, environmental factor ≥ 3) are set based on the expert's empirical consensus on clinical sudden risks and environmental interference sensitivity. After all the superimposed calculations are completed, a drug consumption prediction set covering the specified future period is generated. This set contains the daily consumption quantity arrays of each drug for each department unit, which is marked as the primary prediction signal.

[0035] The primary prediction signal is embodied as a two-dimensional table data structure. The row index is the combined key of the drug code and the department unit path, and the column index is the date sequence of the next seven days. The cell value represents the expected consumption quantity on that day.

[0036] Exemplarily, for the standard code YA002347 of the antihypertensive drug "Amlodipine Besylate Capsules" in the cardiovascular medicine ward, the prescription record sequence of the hypertensive patient group in this department in the past 30 days is extracted by calling the basic drug management data set. It is found that the prescription volume has mutated and increased by 50% in the recent 3 days, and the clinical activity factor value is recorded as 50 at this time. Calculate the change rate of the inventory consumption of this drug in the cardiovascular medicine cabinet. The average slope in 7 days indicates that 3 more boxes are used every day, and the corresponding inventory growth and decline factor value is 300. Access the meteorological data and issue a red high temperature warning for 4 consecutive days, generating 4 environmental interference records with the corresponding factor value of 4. The benchmark prediction of this drug adopts the average daily consumption of 10 boxes last month, and the basic trend extrapolation maintains this value.

[0037] Due to the mutation of the clinical activity factor and the environmental interference factor being greater than 3, the weight of the clinical activity factor is adjusted to 0.75, the weight of the inventory growth and decline factor is 0.10, and the weight of the environmental interference factor is 0.15. Add the benchmark 10 boxes to the results of multiplying the three by their respective weights to get 45 boxes, which is recorded as the predicted value for the next day. After this process traverses all drug department combinations, a 7-day consumption quantity table is output, which is the primary prediction signal.

[0038] In one embodiment of the present invention, step S3 includes the following steps: Obtain the current limited rule library for the safe use of drugs; when a specific drug demand mutation occurs in the primary prediction signal, automatically check the associated medical contraindication information or treatment specifications; if a conflict point is verified, adjust the composition weight of the primary prediction signal and output a corrected reliable demand prediction instruction.

[0039] Specifically, the central storage area loads a preset limited rule library for the safe use of drugs as a verification reference library. The limited rule library for the safe use of drugs consists of an electronic rule list maintained by the hospital pharmacy department and is accessed into the system through a database interface. Its content includes a set of disease classification codes corresponding to drug contraindications, a record of the mapping relationship of pharmacologically equivalent alternative drugs, and a numerical table of the maximum daily payment quantity for each medical insurance category of drugs. Traverse the primary prediction signals generated in step S2 and scan the future one-week consumption quantity array corresponding to the combination of drug codes and department unit paths row by row. Calculate the deviation degree of the predicted consumption quantity on the current day from the average value in the past thirty days for each combination. If the predicted value on a certain day exceeds 50% of the average value, it is marked as a specific drug demand mutation event. The specific drug demand mutation event is set as the fluctuation range of the single-day predicted consumption quantity compared with the historical average value. The initial threshold is fixed at the trigger line of more than 50% increase or decrease, and this threshold can be modified through the system configuration file.

[0040] When this mutation event is detected, immediately retrieve the patient group label associated with this department unit in the basic drug management data set, and query three types of entries in the limited rule library in combination with the standard drug code. They are the list of patient groups with contraindications, the corresponding table of alternative drug relationships, and the medical insurance usage control threshold. The list of patient groups with contraindications specifically associates the standard drug code recorded in the basic drug management data set with the patient group label of the department unit for matching inspection; the corresponding table of alternative drug relationships defines the preferred alternative plan when there is a supply problem with a certain drug; the medical insurance usage control threshold limits the maximum daily prescribing quantity of medical insurance drugs. The conflict points are divided into three levels: If it is retrieved and found that there is a match between the patient group label with contraindications and the drug, a serious conflict point is triggered; If the corresponding table of alternative drug relationships shows that there are recommended alternatives in the near future and the inventory is sufficient, a moderate conflict point is triggered; If the predicted consumption quantity breaks through the medical insurance single-day maximum payment limit, a minor conflict point is triggered.

[0041] Selectively adjust the weight coefficient of the impact factor set according to the conflict level. For severe conflict points, directly remove the clinical activity factor and set its weight value to zero. For moderate conflict points, retain the clinical activity factor but reduce its weight by 0.5 and increase the prediction weight of the alternative drug at the same time. For mild conflict points, maintain the weight but insert the medical insurance limit as the upper threshold to modify the predicted output value. After completing all conflict detections and weight corrections, re-execute the impact factor superposition calculation to overwrite the original primary prediction signal and generate a new two-dimensional table marked as a reliable demand prediction instruction. The output format of the reliable demand prediction instruction is the same as that of the primary prediction signal, but the content is corrected.

[0042] Exemplarily, scanning the primary prediction signal finds that the next-day predicted quantity of the targeted drug "Osimertinib Mesylate" with the standard code YA076521 in the oncology ward suddenly increases to 80 boxes, which is a 167% increase compared to the historical average of 30 boxes, and it is determined as a demand mutation. Retrieving the limited rule library obtains three relevant rules: the contraindication list requires that patients with negative EGFR genes are prohibited from using it, the alternative relationship table shows that the optional alternative drug is "Gefitinib", and the medical insurance regulation limits the daily prescription to 10 boxes. Matching the basic drug management data set shows that the current patient group label in this department includes 15 "EGFR-negative lung cancer patient groups", triggering a first-level conflict point. The system executes the severe conflict handling process: reduces the weight of the clinical activity factor to 0, increases the weight of the inventory growth and decline factor to 0.8, and the weight of the environmental interference factor to 0.2. Recalculate the predicted quantity of this drug according to the new weights. The original benchmark quantity of 20 boxes is superimposed with the inventory growth and decline factor value of 50 boxes and the environmental interference factor value of 8 boxes. The calculation process is 20 boxes + (50 × 0.8) + (8 × 0.2) = 42 boxes. Compare 42 boxes with the medical insurance limit of 10 boxes and take the smaller value. Finally, the next-day consumption quantity of this drug in the reliable demand prediction instruction is output as 10 boxes. This correction process synchronously triggers an instruction to increase the prediction weight of the alternative drug Gefitinib.

[0043] In one embodiment of the present invention, step S4 includes the following steps: Continuously monitor the batch inventory status of each pharmacy node, and calculate the quantitative evaluation value reflecting the drug turnover efficiency, expiration pressure, and cost loss; when the quantitative evaluation value exceeds the safety warning threshold, automatically generate an inventory exception warning signal including the risk level and the associated pharmacy.

[0044] Specifically, poll the basic drug management data set daily to obtain the current batch inventory status of each pharmacy node. The batch inventory status points to the physical inventory unit recorded in the basic drug management data set. Each unit includes the unique identifier of the drug code, the physical storage location code, the current inventory quantity of the goods, the batch lot number string, and the information of the effective expiration date.

[0045] For the inventory pool of each drug code in each pharmacy, perform three parallel calculation operations: 1. Drug turnover efficiency is an objective value that quantifies the effectiveness of inventory turnover. This is achieved by taking the total amount of drugs shipped out during the seven-day period preceding the current monitoring date, dividing it by the total real-time inventory on that monitoring date, multiplying by 100, and rounding the result to the nearest integer. A higher value indicates faster turnover and lower risk. 2. Expiration pressure is achieved by scoring the backlog level of near-expiry drugs, identifying the total inventory of expired batches within 30 days and converting it into a percentage of the total inventory of the drug; 3. Cost loss is set as the waste risk prediction of high-value drugs. The 20% drug catalog with the highest unit price is read from the hospital cost system. The near-expiry pressure value of each variety in the catalog is multiplied by the unit price of the drug, and then the results of each variety are accumulated.

[0046] The calculation of the quantitative evaluation value satisfies the following formula:

[0047] in, represents the comprehensive risk score; An indicator that indicates drug turnover efficiency; Indicators of peri-term stress; An indicator that represents cost loss; It is the benchmark value of the maximum possible loss in the entire hospital obtained from historical data statistics. The weight coefficient is set based on the priority balance of the core objectives of comprehensive drug management (ensuring supply security, reducing shelf life loss, and controlling costs), and the weight of each indicator's impact on actual losses is determined through historical risk event attribution analysis, and is confirmed by experts from the Department of Pharmacy, the Finance Department, and the Clinical Department. This reflects the strong correlation between turnover efficiency and drug discontinuation risk. Corresponding to the clinical quality control requirements for scrapping of drugs nearing their expiration date, Reflects the level of refinement in cost control of high-value drugs.

[0048] If the comprehensive evaluation value exceeds the safety warning threshold preset according to the pharmacy type, the risk judgment engine is triggered. The setting basis of the safety warning threshold includes the pharmacy type differentiated by the pharmacy of the third-level first-class hospital and the community pharmacy, the drug classification characteristics differentiated by the emergency medicine and chronic disease medicine, and the upper limit of the comprehensive risk acceptable to each pharmacy through the GSP regulations, which is calculated by the product of the daily consumption volume, the procurement cycle, and the safety factor. For example, the outpatient pharmacy is set to 60, the inpatient pharmacy is set to 40, and the emergency pharmacy is set to 80. The risk judgment engine first checks the expected consumption trend of the relevant drugs in the pharmacy according to the reliable demand prediction instruction. If the expected consumption volume is lower than 80% of the daily average volume for three consecutive days, the risk level is raised. The risk level classification standard is the percentage by which the comprehensive evaluation value exceeds the safety warning threshold. Exceeding 0-30% (including 30%) is a low-level risk, exceeding 30%-70% (including 70%) is a medium-level risk, and exceeding 70% is a high-level risk.

[0049] Finally, an inventory exception warning signal in string structure is automatically generated, defined as a JSON format message. The message header contains the drug code and the pharmacy path code to which it belongs. The middle part marks the high, medium, and low risk labels, and the tail part adds a description field of the main risk source.

[0050] In one embodiment of the present invention, step S5 includes the following steps: Receive the reliable demand prediction instruction and the inventory exception warning signal, and dynamically generate a cross-warehouse transfer plan that simultaneously meets the following objectives according to the preset cost priority level and service guarantee requirements, match the future demand and maintain a reasonable safety inventory; give priority to processing high-risk near-expired drugs; minimize the frequency of emergency transfers, and output a transfer strategy instruction including the drug flow direction, the allocation quantity, and the timeliness.

[0051] Specifically, a central decision-making engine is established to access the reliable demand prediction instruction and the inventory exception warning signal that is refreshed in real time. The central decision-making engine is configured with an independently running service process, and listens to the change event of the reliable demand prediction instruction data table and the push event of the inventory exception warning signal through a message queue; the inventory exception warning signal queue is implemented as a first-in-first-out data structure, and each record is sorted according to the received timestamp and continuously input into the engine for processing. When a new warning signal or a reliable prediction instruction update is detected, the scheduling operation is triggered.

[0052] Initialize the temporary area for drug allocation strategies and perform cross-library retrieval operations on each drug code with inventory anomalies: In the first step, traverse the total available inventory of all pharmacy nodes; in the second step, match the one-week predicted consumption array of the drug in the relevant departments in the reliable demand forecasting instructions; in the third step, calculate the safety stock baseline value by taking three times the daily average consumption of the department. The cross-library retrieval operation refers to the distributed query process of the basic drug management dataset, and each retrieval obtains the real-time total inventory and valid batch details of the drug in all hospital pharmacy nodes. The safety stock baseline value setting rule is the median of the daily average consumption data sample of the department unit in the past three months multiplied by a three-fold buffer coefficient.

[0053] Implement the inventory rebalancing priority strategy for high-risk near-expiry drugs, that is, transfer the available inventory from the pharmacy node with the highest expiration risk to replenish the safety stock gap in the pharmacy node. If the inventory of a high-risk drug in the source pharmacy can cover the consumption of the target pharmacy for more than one week, a point-to-point direct transfer plan is generated; otherwise, start the drug centralized redistribution process. High-risk near-expiry drugs specifically refer to drug codes marked as high-risk level in the inventory anomaly warning signal and whose dominant risk source includes near-expiry pressure; the inventory rebalancing priority strategy defines the allocation order logic, always giving priority to handling the backlog inventory of high-risk drugs. The allocation process needs to ensure that the remaining stock in the source node is not less than the three-day consumption.

[0054] The point-to-point direct transfer plan is applicable to single-source and single-target scenarios, and the scheduling volume takes the safety stock gap value of the target node but does not exceed the upper limit of the available inventory that can be released by the source node; the drug centralized redistribution process is used for multi-source or multi-target scenarios, and the coordinated central pharmacy serves as a transit node to accept the inventory of each high-risk source and redistribute it to the gap nodes again. The upper limit of the available inventory that can be released is set based on the remaining amount after deducting its safety stock threshold (including the buffer for in-transit orders) from the real-time total inventory of the drug in the source node, and is superimposed with the priority release of near-expiry batches and ensuring the dynamic calibration of the basic demand supply of this node in the next week.

[0055] For example, the upper limit of the available inventory that can be released is set as follows: The total inventory of insulin in a certain pharmacy is 200 units, the safety stock threshold is 80 units (meeting the usage of 10 units per day for 7 days), and the near-expiry batch is 100 units, then the maximum amount that can be released is 120 units; however, if 60 units have a shelf life of less than 7 days, the actual amount that can be released needs to be adjusted to 60 units to ensure that 140 units of effective inventory are still retained in this pharmacy after release to cover the safety threshold and shelf life buffer.

[0056] Each allocation operation calculates three cost factors cumulatively: The in-transit loss coefficient of the drug takes 1% of the near-expiry pressure value, the cumulative penalty term for the allocation frequency takes the number of times scheduled this month multiplied by the fixed cost per operation, and the value factor of the unit distance distribution cost is set according to the logistics contract at the benchmark price per kilometer.

[0057] Select the optimal plan using the lowest total score method, satisfying the following formula:

[0058] in, Indicates the total score of the scheduling plan; Indicates the allocation quantity, which is the target pharmacy's safety stock gap; Indicates the drug in-transit loss coefficient; Indicates the cumulative penalty item for transfer frequency; Represents the distribution cost per unit distance value factor. If the total score of a scheduling plan is 20% lower than the historical best value, it will be adopted. Dynamically output the allocation strategy instruction set, which includes the drug standard code, source location code, destination location code, allocation quantity integer, and urgency label.

[0059] The allocation strategy instruction set is output as a structured instruction sequence. Each instruction contains a sixteen-digit drug standard code string, a code string for the outgoing pharmacy node, a code string for the incoming pharmacy node, an integer value for the delivery quantity, and an urgency field. At the same time, ordinary-level drugs must be delivered within twenty-four hours, and emergency-level drugs must be delivered within eight hours.

[0060] For example, a received inventory anomaly alarm signal numbered A1875 indicates that the total inventory of the antiemetic "Aprepitant Capsules," standard code YA088216, in the oncology day ward's medicine cabinet is 100 pills, with a 60% expiration risk, indicating a high risk level. The central decision engine searches the hospital's inventory and finds 500 pills in the central drug store and 300 in the west pharmacy. Matching reliable demand forecasts indicates that the oncology day ward's average daily consumption for the next seven days is 20 pills. Using a safety stock baseline of 60 pills, the current shortfall is 40 pills.

[0061] Triggering the inventory rebalancing priority strategy (high-risk drug): Calculating the central pharmacy's expiring pressure is only 10%, select it as the source for transfers, and generate a direct transfer plan for 40 pills to the day ward. Calculating the cost factor yields a 0.6 coefficient for expiring pressure loss, a distance factor of 2 kilometers, and a penalty of 0.2 yuan for the third transfer this month. The dispatch plan's total score is 44.4, 44% lower than the historical average of 80. The output command is automatically adopted: the central pharmacy code for the transfer is PH001, the code for the transferee is PH205, the quantity of drug YA088216 is 40 pills, and the delivery time is 8 hours, with an emergency level.

[0062] If a new alarm is received at this time showing that the same drug is also at high risk of expiration in the emergency pharmacy, the redistribution process will be initiated: the central drug warehouse will release 100 pills first to the transit warehouse, and then divided into 40 pills for the day ward and 60 pills for the emergency department. The plan with the lowest total score will have higher priority than single-point allocation.

[0063] In one embodiment of the present invention, step S6 includes the following steps: Obtain the actual drug consumption progress and scheduling task completion records, and compare the deviation values between the reliable demand prediction instructions, allocation strategy instructions and actual results; if the deviation value indicates that the judgment of the original limited rule base fails or there is a new risk mode, prompt to update the content of the limited rule base for the safe use of the drug.

[0064] Specifically, grab the distribution receipt records of the executed allocation strategy instructions through the hospital logistics system interface and correlate the pharmacy consumption report data to construct an execution tracking data set. Poll this data set daily and perform triple comparison operations on the dimension of each department unit belonging to the drug standard code: The first comparison is to calculate the daily average consumption deviation rate between the one-week expected consumption array in the reliable demand prediction instruction and the actual pharmacy consumption record; the actual drug consumption progress is the net consumption value of the daily manual confirmation or the automatic reported sales out minus return quantity of each pharmacy node. The second verification is the difference value between the planned allocation quantity and the actual inbound quantity of the allocation strategy instruction. The scheduling task completion record is taken from the electronic handover document of the hospital internal logistics system, including the distribution order number, drug code, shipping pharmacy, receiving pharmacy, actual delivered quantity, and receipt time fields. The third check is the change range of the turnover efficiency after the disposal of the inventory exception warning signal corresponding to high-risk drugs.

[0065] The deviation value between the reliable demand prediction instruction, the allocation strategy instruction and the actual result is that the consumption deviation rate is the percentage of the absolute value of the difference between the actual daily average consumption minus the predicted daily average consumption divided by the predicted value; the allocation difference value takes the absolute value of the planned inbound quantity minus the actual inbound quantity; the change range of the turnover efficiency is defined as the result of subtracting the original value from the turnover efficiency index after disposal. When any dimension exceeds the set deviation threshold for three consecutive days, start the rule failure analysis process to extract the key constraint entries of the drug in the limited rule base and trace back the execution process data.

[0066] The deviation threshold is based on the deviation distribution characteristics between prediction and actual in historical operation data (such as taking the 90th percentile value of the consumption deviation rate as the benchmark), superimposing the drug criticality classification (such as tightening the threshold for emergency drugs) and management granularity requirements (such as stricter monitoring for high-value drugs), and finally determining the sensitivity of the threshold to business risks through Monte Carlo simulation. For example, the consumption deviation rate threshold is set at 15%, because the predicted deviation of 90% of the conventional drugs in historical data < 12%, leaving a 3% buffer; the allocation difference threshold of 5 boxes corresponds to the logistics error control standard; the change range of the turnover efficiency of ±10% is the recognized inventory health fluctuation boundary of the pharmacy department.

[0067] For the case where the daily average consumption deviation rate exceeds the consumption deviation rate threshold, check whether the matching of the original medical contraindication information is accurate; for the scenario where the dispensing difference value exceeds the dispensing difference threshold and the emergency dispatch flag is set, verify the rationality of the distribution time limit rule; for the cases where the change range of the turnover efficiency fails to reach the expected improvement, verify the calculation logic of the approaching expiration pressure. If the analysis determines that there is a lack of entries in the defined rule library or the parameters are outdated, generate a structured revision suggestion and push it to the review interface of the pharmacy department. The suggested content includes the original rule identifier, the revision type code, and the adjusted value of the influencing factor. The new risk mode refers to the situation where high-frequency deviations are found in the execution tracking dataset but there are no relevant constraint entries in the defined rule library. The mode characteristics are manifested as the same anomalies occurring for the same type of drugs in multiple departments.

[0068] The content update process of the defined rule library for the safe use of drugs stipulates that the revised form needs to be electronically signed by the person in charge of pharmacy and then transmitted back to the system, which will automatically synchronize to the central storage area to overwrite the original entry. The prompt update operation is designed to generate a standard revised form by the system. The form includes the triggering drug code, the invalidation rule number, the revision type field, with options of addition, deletion, and correction, the suggested value of the revised parameter, and a text area for the revision description.

[0069] Exemplarily, tracking the execution data of the standard code YA099182 of the targeted drug trastuzumab in the oncology ward, the reliable demand prediction instruction shows that 10 doses should be consumed on the seventh day, but actually only 5 doses are consumed, and the consumption deviation rate is 50%, exceeding the threshold for three consecutive days. Checking the original defined rule library entry R803, which prohibits use in HER2-negative patients, and matching the drug management basic dataset, it is found that three new HER2-negative patients have been added in this department, triggering the rule invalidation process. After analysis and confirmation, this deviation is caused by the misjudgment rate of the new gene detection technology, and the system generates a revised suggestion form: Mark the drug YA099182 with the rule number R803, select the correction type, modify the parameter suggested value, and add a supplementary clause that if the patient has a PS3 gene mutation, the use restriction is waived. At the same time, it is found that a similar deviation occurs for the same drug in the gynecology ward, but there is no relevant constraint in the defined rule library, which is determined as a new risk mode. Add a new rule entry R831 stating the reduction criteria for the auxiliary treatment cycle of breast cancer. After the two suggestions are approved by the pharmacy director, they take effect immediately, and thereafter similar deviations disappear, and the system completes the closed-loop optimization.

[0070] See Appendix Figure 2 , the present invention also proposes an artificial intelligence-based hospital comprehensive monitoring and analysis platform, including the following modules: A multi-source medical data collection module, which is used to obtain the operation records, inventory status, medical behaviors, and external environment indicators involved in the drug circulation in the hospital, perform structured cleaning according to the drug types, department units, and patient group types, and generate a drug management basic dataset; A drug demand prediction module, which is used to identify the characteristics of clinical diagnosis and treatment activities, the growth and decline rules of pharmacy inventory, and environmental interference factors based on the basic drug management dataset, calculate the expected consumption of various drugs in each department unit within a specified future period, and generate a primary prediction signal; A clinical rule verification module, which is used to call the drug safe use limit rule library for clinical rationality verification when a specific drug demand mutation is detected in the primary prediction signal, adjust the prediction weight according to the conflict level, and output a reliable demand prediction instruction; An inventory risk scanning module, which calculates a quantitative evaluation value reflecting drug turnover efficiency, expiration pressure, and cost loss by continuously monitoring the batch inventory status of each pharmacy node, and generates an inventory exception warning signal when the quantitative evaluation value exceeds the preset safety warning threshold; A multi-objective scheduling decision module, which is used to receive the reliable demand prediction instruction and the inventory exception warning signal, dynamically generate a cross-library transfer plan according to the preset cost priority and service guarantee requirements, and output a transfer strategy instruction including drug flow direction, allocation quantity, and timeliness; An execution feedback analysis module, which obtains the actual drug consumption progress and the scheduling task completion record, compares the deviation value between the reliable demand prediction instruction and the transfer strategy instruction, and triggers an update prompt for the drug safe use limit rule library when the deviation value continuously exceeds the preset deviation threshold.

[0071] It should be noted that for the formulas mentioned above, through the principle of dimensional consistency and mathematical standardization means (such as normalization processing, dimensionless parameter conversion, or unit system unification), physical quantities with different attributes can be translated into dimensionless standard values or superposable parameters with the same dimension, so as to eliminate the interference of different dimensions on the operation logic, and make the formulas have mathematical operation rationality and objective law adaptability while retaining the original data distribution characteristics. The above is only an exemplary embodiment of the present invention, and the scope of the present invention cannot be limited thereby.

[0072] Each of the above-mentioned modules can be implemented in whole or in part by software, hardware, and their combination, supported in the form of hardware embedded in or independent of the processor in a computer device, and also supported in the form of software stored in the memory in a computer device, facilitating the processor to call and execute the operations corresponding to each of the above-mentioned modules.

[0073] It should be noted that the human body information (including but not limited to human device information and personal information, etc.) and data (including but not limited to data for analysis, stored data, and displayed data, etc.) involved in the present invention are all information and data authorized by the human body or fully authorized by all parties. The collection, use, and processing of relevant data require relevant legal standards.

[0074] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An integrated hospital monitoring and analysis method based on artificial intelligence, characterized in that, It includes the following steps: S1. Obtain the operation records, inventory status, medical behaviors, and external environment indicators involved in the hospital drug circulation, perform structured cleaning according to drug types, department units, and patient group types, and generate a basic drug management dataset; S2. Based on the basic drug management dataset, identify the characteristics of clinical diagnosis and treatment activities, the growth and decline rules of pharmacy inventory, and environmental interference factors, calculate the expected consumption of various drugs in each department unit within a specified future period, and generate a primary prediction signal; S3. When a specific drug demand mutation is detected in the primary prediction signal, call the drug safe use limitation rule library for clinical rationality verification, adjust the prediction weight according to the conflict level, and output a reliable demand prediction instruction; S4. By continuously monitoring the batch inventory status of each pharmacy node, calculate the quantitative evaluation values reflecting drug turnover efficiency, approaching expiration pressure, and cost loss. When the quantitative evaluation value exceeds the preset safety warning threshold, generate an inventory exception warning signal; S5. Receive the reliable demand prediction instruction and the inventory exception warning signal, dynamically generate a cross-library transfer plan according to the preset cost priority and service guarantee requirements, and output a transfer strategy instruction including drug flow direction, allocation quantity, and timeliness; S6. Obtain the actual drug consumption progress and the completion record of the scheduling task, compare the deviation value between the reliable demand prediction instruction and the transfer strategy instruction. When the deviation value continuously exceeds the preset deviation threshold, trigger an update prompt for the drug safe use limitation rule library.

2. The hospital comprehensive monitoring and analysis method based on artificial intelligence according to claim 1, characterized in that Generate a basic drug management dataset, including the following steps: Collect the original data stream through a preset periodic scraping mechanism. The original data stream covers the operation records, inventory status, medical behaviors, and external environment indicators involved in the hospital drug circulation, and perform classification and completion operations on the missing fields; Unify the timestamp conversion to the format of year, month, day, and hour, and map the free-text description of the drug administration method to a standard operation code; The converted original data stream is classified and reorganized according to three types of indexes: drug types, department units, and patient group types. The data under the same type of index is arranged in chronological order to generate the columnar storage basic drug management dataset.

3. The method for comprehensive hospital monitoring and analysis based on artificial intelligence according to claim 2, characterized in that The converted original data stream is classified and reorganized according to three types of indexes: drug types, department units, and patient group types, including the following steps: The drug type dimension points to the unique drug identification identifier established based on the pharmacological classification code and dosage form specifications; The department unit dimension includes the outpatient pharmacy window number, complete hospital area and building path information; The patient group type dimension is configured as a set of labels divided according to disease characteristics, and the label generation logic is defined as automatically marked if the patient's current main diagnosis code belongs to the cardiovascular disease classification catalog or the age is within the pediatric admission range.

4. The method for comprehensive hospital monitoring and analysis based on artificial intelligence according to claim 2, wherein, Generate a primary prediction signal, including the following steps: Extract the characteristics of clinical diagnosis and treatment activities, the growth and decline rules of pharmacy inventory, and environmental interference factors from the basic drug management dataset, and convert them into a set of numerical impact factors; Adopt basic trend extrapolation to calculate the benchmark prediction quantity of each drug in the department unit; According to the preset dynamic weight distribution principle, superimpose the set of numerical impact factors on the benchmark prediction quantity to generate a primary prediction signal including an array of future multi-day consumption quantities; The primary prediction signal is embodied as a two-dimensional table data structure. The row index is the combined key of the drug code and the department unit path, the column index is the date sequence of the next seven days, and the cell value represents the expected consumption on that day.

5. The hospital comprehensive monitoring and analysis method based on artificial intelligence according to claim 2, characterized in that, Call the drug safety use limitation rule library for clinical rationality verification, including the following steps: When the predicted consumption for a single day exceeds the preset fluctuation threshold compared with the historical average, it is determined as a specific drug demand mutation event; Associate the patient group label in the basic drug management data set with the list of contraindicated patient groups, the corresponding table of alternative drug relationships, and the medical insurance usage control threshold in the limitation rule library for conflict detection; According to the levels of severe conflict points, moderate conflict points, or mild conflict points, selectively adjust the weight coefficients of clinical activity factors, inventory growth and decline factors, and environmental interference factors; The limitation rule library consists of an electronic rule list, including the set of disease classification codes corresponding to drug contraindications, the record of pharmacological equivalent alternative drug mapping relationships, and the table of the maximum daily payment quantity values for drugs of each medical insurance category.

6. The method for comprehensive hospital monitoring and analysis based on artificial intelligence according to claim 2, wherein, Generate an inventory exception warning signal, including the following steps: Calculate drug turnover efficiency indicators, approaching expiration pressure indicators, and cost loss indicators; Synthesize a comprehensive risk score according to the preset weight coefficients, and the weight coefficients are determined based on the attribution analysis of historical risk events for the core objectives of drug management; When the comprehensive risk score exceeds the preset safety warning threshold associated with the pharmacy type, an inventory exception warning signal in JSON format is generated with an additional risk level label. The header of the inventory exception warning signal message contains the drug code and the pharmacy path code to which it belongs, the risk label is marked in the middle, and the description field of the main risk source is appended at the end.

7. An integrated hospital monitoring and analysis method based on artificial intelligence according to claim 2, characterized in that, Dynamically generate a cross-library transfer plan according to the preset cost priority and service guarantee requirements, including the following steps: Perform a cross-library retrieval operation on the drugs with inventory exceptions, obtain the total available inventory in the whole hospital and match the safety inventory baseline value; Start the inventory rebalancing priority strategy for high-risk approaching expiration drugs, and generate a point-to-point direct transfer plan or a drug centralized redistribution process according to the upper limit of the available inventory that can be released from the source pharmacy; Use the total score lowest method to calculate the total score of the scheduling plan, and the total score of the scheduling plan is dynamically synthesized by the transfer quantity, the in-transit loss coefficient of the drug, the cumulative penalty item of the transfer frequency, and the unit distance distribution cost value factor.

8. An integrated hospital monitoring and analysis method based on artificial intelligence according to claim 7, characterized in that, Generate a point-to-point direct transfer plan or a drug centralized redistribution process according to the upper limit of the available inventory that can be released from the source pharmacy, including the following steps: The point-to-point direct transfer plan is applicable to the single-source single-target scenario, and the transfer quantity takes the safety inventory gap value of the target node but does not exceed the upper limit of the available inventory that can be released from the source node; The drug centralized redistribution process is used for multi-source or multi-target scenarios, and the coordinated central pharmacy serves as a transfer node to accept the re-secondary distribution of the inventory of each high-risk source to the gap nodes; The setting of the upper limit of the available inventory that can be released is based on the remaining quantity after deducting the safety inventory threshold from the real-time total inventory of the drugs at the source node, and is superimposed with the priority release of the approaching expiration batches and the dynamic calibration of ensuring the basic demand supply of this node in the next week.

9. The method for comprehensive hospital monitoring and analysis based on artificial intelligence according to claim 2, wherein, When the deviation value continuously exceeds the preset deviation threshold, an update prompt for the drug safety use limitation rule library is triggered, including the following steps: Calculate three types of deviation indicators: the daily average consumption deviation rate, the dispensing difference value, and the change range of turnover efficiency. When any deviation index exceeds the corresponding preset deviation threshold for N consecutive days, trace back the logical association between the entries in the rule library and the execution process data; If a rule is detected to be missing or a parameter is outdated, generate a structured revision recommendation containing a revision type code and an impact factor adjustment value; The deviation threshold is finally determined by superimposing the drug criticality classification and management granularity requirements based on the deviation distribution characteristics between the prediction and the actual situation in the historical operation data and verifying the sensitivity of the threshold to business risks through simulation.

10. An integrated hospital monitoring and analysis platform based on artificial intelligence, characterized in that, It includes the following modules: A multi-source medical data collection module, which is used to obtain the operation records, inventory status, medical behaviors, and external environment indicators involved in the hospital drug circulation, perform structured cleaning according to drug types, department units, and patient group types, and generate a basic drug management data set; A drug demand prediction module, which is used to identify the characteristics of clinical diagnosis and treatment activities, the law of pharmacy inventory rise and fall, and environmental interference factors based on the basic drug management data set, calculate the expected consumption of various drugs in each department unit within a specified future period, and generate a primary prediction signal; A clinical rule verification module, which is used to call the drug safe use limit rule library for clinical rationality verification when a specific drug demand mutation is detected in the primary prediction signal, adjust the prediction weight according to the conflict level, and output a reliable demand prediction instruction; An inventory risk scanning module, which calculates a quantitative evaluation value reflecting drug turnover efficiency, approaching expiration pressure, and cost loss by continuously monitoring the batch inventory status of each pharmacy node, and generates an inventory exception warning signal when the quantitative evaluation value exceeds the preset safety warning threshold; A multi-objective scheduling decision-making module, which is used to receive reliable demand prediction instructions and inventory exception warning signals, dynamically generate a cross-warehouse transfer plan according to the preset cost priority and service guarantee requirements, and output a transfer strategy instruction containing the drug flow direction, allocation quantity, and timeliness; An execution feedback analysis module, which obtains the actual drug consumption progress and the completion record of the scheduling task, compares the deviation value between the reliable demand prediction instruction and the transfer strategy instruction, and triggers an update prompt for the drug safe use limit rule library when the deviation value continuously exceeds the preset deviation threshold.

Images