Data monitoring method and device for medicine collection channel

Abstract

The invention discloses a data monitoring method and device oriented to a medicine collection channel, and relates to the field of medical data detection, and the method comprises the steps: setting a policy event time window of a target medicine, collecting daily average quotation data in the window, carrying out the standardization processing of the data based on a mean value and a standard deviation of a previous price stabilization period of a policy, and obtaining a standard quotation value; generating a standardized sequence; inputting the sequence into an anomaly detection algorithm to identify price anomaly points, calculating advanced response time, a supplier synchronous response index and a standardized price of each anomaly time point, and constructing an anomaly score; and when the score exceeds a preset threshold value, the system outputs early-warning information about abnormal advanced quotation. The medical price monitoring system effectively makes up the defect of an existing medical price monitoring system in the aspect of policy awareness, and has the value of early warning for market prices for medical insurance institutions, supervision departments, collection platforms, purchasing departments of hospitals and the like.

Description

Technical Field

[0001] The present invention relates to the field of medical data detection, and more particularly to a data monitoring method and device for medicine collection channels. Background Art

[0002] During the preparatory and negotiation phase before the official release of a policy, pharmaceutical suppliers often preemptively respond to its bidding strategies based on industry expectations and non-public information, aiming to mitigate bid risks, lock in orders in advance, or implement tentative price reductions. This behavior has led to price fluctuations for some drugs before the policy is implemented, creating the phenomenon of "preemptive pricing anomalies during policy-sensitive periods."

[0003] In actual data collection channels, this type of abnormal early quote behavior typically lacks global characteristics, but rather manifests as sudden fluctuations within a localized time period, often accompanied by synchronized reactions from multiple suppliers. Traditional price fluctuation detection methods often focus on overall price trends, long-term mean or variance changes, and are unable to capture these short-term, low-frequency, and coordinated hidden abnormal signals.

[0004] Therefore, there is an urgent need to conduct anomaly detection in the window period before policy release. Based on the statistical characteristics of drug prices in a stable period, combined with event-driven time window settings, through standardized processing, sliding window anomaly detection and multi-supplier response feature extraction mechanisms, automatic identification and risk scoring of abnormal behaviors in advance quotations can be achieved, thereby providing regulatory agencies, centralized procurement platforms, medical institutions, hospital procurement departments, etc. with highly sensitive and low-cost abnormal behavior warning tools. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a data monitoring method and device for medicine collection channels to solve the problems mentioned in the background technology.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A data monitoring method for pharmaceutical collection channels, comprising the following steps:

[0008] Setting a policy event time window for the target drug, where the policy event time window includes a period of time before the policy issuance date;

[0009] Collecting daily average quotation data of target drugs within the policy event time window;

[0010] Normalizing the daily average quotation data based on the mean and standard deviation of a period of price stability before the policy event time window to generate standardized series data reflecting the degree of volatility;

[0011] Inputting the standardized sequence data into an anomaly detection algorithm to identify price anomaly point data appearing in the policy event time window and forming a corresponding anomaly time point set;

[0012] For each price anomaly point data, calculate the lead time from the corresponding abnormal time point date to the policy release date, calculate the synchronous response index of multiple suppliers at the same abnormal time point date, and the price in the standardized series data at the same abnormal time point date;

[0013] constructing an anomaly score based on the lead response time, the synchronous response index, and prices in the standardized sequence data;

[0014] When the abnormality score exceeds a preset threshold, early abnormal quotation warning information for the target drug is output.

[0015] In some embodiments, the normalization process corresponds to the formula:

[0016] where Z t is the standardized sequence data that changes with time t after standardization, P t is the average daily quote, μ is the mean of the price during the price stability period, and σ is the standard deviation of the price during the price stability period.

[0017] In some embodiments, the anomaly detection algorithm is a cumulative sum detection algorithm CUSUM.

[0018] In some embodiments, the formula for the abnormality score is:

[0019]

[0020] in:

[0021] S t is the anomaly score corresponding to the abnormal time point date t;

[0022] T is the policy issuance date;

[0023] R t is the supplier's synchronous response index at the abnormal time point date t;

[0024] Z t is the standardized price;

[0025] σ is the standard deviation of prices during the period of price stability;

[0026] w1, w2, w3 are the corresponding weight coefficients.

[0027] In some embodiments, R t The calculation formula is:

[0028]

[0029] Where N2(t) is the number of all suppliers who submitted quotations at the abnormal time point date t;

[0030] N1(t) is the number of suppliers whose absolute value of the difference between their quotations and the mean μ at the abnormal time point t exceeds the threshold δ.

[0031] In some embodiments, the threshold value is calculated as follows: Among them, μ is the mean of the price during the price stability period, σ is the standard deviation of the price during the price stability period, and α is the adjustment coefficient.

[0032] The present invention also discloses a data monitoring device for medicine collection channels, comprising:

[0033] An event window setting unit, configured to set a policy event time window for a target drug, wherein the policy event time window includes a period of time before the policy release date;

[0034] a quotation data collection unit, configured to collect daily average quotation data of target drugs within the policy event time window;

[0035] a normalization processing unit, configured to perform normalization processing on the daily average quotation data based on the mean and standard deviation of a period of price stability before the policy event time window, to generate normalized series data reflecting the degree of volatility;

[0036] an anomaly detection unit, configured to input the standardized sequence data into an anomaly detection algorithm, identify price anomaly point data appearing in the policy event time window, and form a corresponding anomaly time point set;

[0037] An indicator calculation unit is used to calculate, for each price anomaly data point, the lead time from the corresponding anomaly time point date to the policy release date, calculate the synchronous response index of multiple suppliers at the same anomaly time point date, and the price in the standardized series data at the same anomaly time point date;

[0038] An early warning output unit is used to construct an abnormality score based on the advance response time, the synchronous response index and the price in the standardized sequence data, and when the abnormality score exceeds a preset threshold, output early abnormal quotation warning information for the target drug.

[0039] In some embodiments, the normalization process corresponds to the formula:

[0040] where Z tis the standardized sequence data that changes with time t after standardization, P t is the average daily quote, μ is the mean of the price during the price stability period, and σ is the standard deviation of the price during the price stability period.

[0041] In some embodiments, the formula for the abnormality score is:

[0042]

[0043] in:

[0044] S t is the anomaly score corresponding to the abnormal time point date t;

[0045] T is the policy issuance date;

[0046] R t is the supplier's synchronous response index at the abnormal time point date t;

[0047] Z t is the standardized price;

[0048] σ is the standard deviation of prices during the period of price stability;

[0049] w1, w2, w3 are the corresponding weight coefficients.

[0050] In some embodiments, R t The calculation formula is:

[0051]

[0052] Where N2(t) is the number of all suppliers who submitted quotations at the abnormal time point date t;

[0053] N1(t) is the number of suppliers whose absolute value of the change in quotation exceeds the threshold δ at the abnormal time point date t.

[0054] The advantages of the present invention over the prior art are:

[0055] 1. The present invention can achieve early warning based on policy events. More specifically, the present invention introduces a "policy event time window" mechanism, focusing on a critical period before the policy is officially released, and combining the fluctuation trend of quotation data within this window to achieve targeted monitoring of policy-driven market behavior, significantly enhancing the timeliness and foresight of quotation anomaly identification, and avoiding the lag of traditional monitoring methods.

[0056] 2. This invention improves the robustness and sensitivity of anomaly identification. More specifically, by introducing the mean and standard deviation of the "price stability period" as a normalization benchmark, combined with the Z-score (normalization processing) mechanism, this invention can effectively eliminate errors caused by natural market fluctuations and improve the ability to identify minor but significant price anomalies, thereby avoiding the disadvantages of both false positives and false negatives.

[0057] 3. This invention establishes a multi-dimensional comprehensive scoring mechanism for higher accuracy. More specifically, this invention innovatively integrates multiple factors, such as lead response time, supplier synchronization response index, and standardized price, to construct a comprehensive anomaly scoring function. When the score exceeds a preset threshold, an alert is triggered. This method is more discriminative and adjustable than traditional single-indicator monitoring methods, facilitating the precise screening of drug pricing behaviors that pose systemic risks.

[0058] 4. This invention can reflect market linkage behavior and reveal potential collusion or trend signals. More specifically, this invention proposes the concept of a "supplier synchronization response index," which quantifies the degree of consistency in quote adjustments across multiple suppliers at the same point in time. This reveals possible market coordination or collective fluctuations driven by the same external signal, which are often overlooked by traditional drug price monitoring methods.

[0059] In summary, the present invention effectively makes up for the shortcomings of existing pharmaceutical price monitoring systems in terms of policy perception, and has the value of providing early warning of market prices to medical insurance institutions, regulatory departments, centralized procurement platforms, hospital purchasing departments, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is an overall flow chart of the method of the present invention;

[0061] Figure 2 is a flow chart of the standardization process of the present invention;

[0062] Figure 3 is a flow chart of outlier identification of the present invention;

[0063] Figure 4 It is a schematic diagram of the present invention for calculating scores and early warnings. DETAILED DESCRIPTION

[0064] The specific embodiments of the present invention will be described below with reference to the accompanying drawings.

[0065] This method systematically analyzes price quotes for target drugs under the influence of policy events, enabling early identification of unusual fluctuations and generating early warning information, providing decision support for regulatory agencies and pharmaceutical procurement platforms. The entire process encompasses time window setting, data collection, standardization, anomaly detection, anomaly score calculation, and early warning output, which will be explained in detail below.

[0066] like Figure 1 As shown, the present invention provides a data monitoring method for pharmaceutical collection channels, which generally includes the following steps:

[0067] Setting a policy event time window for the target drug, where the policy event time window includes a period of time before the policy issuance date;

[0068] Collecting daily average quotation data of target drugs within the policy event time window;

[0069] Normalizing the daily average quotation data based on the mean and standard deviation of a period of price stability before the policy event time window to generate standardized series data reflecting the degree of volatility;

[0070] Inputting the standardized sequence data into an anomaly detection algorithm to identify price anomaly point data appearing in the policy event time window and forming a corresponding anomaly time point set;

[0071] For each price anomaly point data, calculate the lead time from the corresponding abnormal time point date to the policy release date, calculate the synchronous response index of multiple suppliers at the same abnormal time point date, and the price in the standardized series data at the same abnormal time point date;

[0072] constructing an anomaly score based on the lead response time, the synchronous response index, and prices in the standardized sequence data;

[0073] When the abnormality score exceeds a preset threshold, early abnormal quotation warning information for the target drug is output.

[0074] In specific embodiments, the policy event time window refers to a specific period of time before the policy release date, used to capture market reactions to policy expectations. To appropriately set the length of the time window, historical data or industry experience can be used as a reference. Specific adjustments can be made based on statistical analysis of historical market reactions or expert judgment of the policy's impact cycle. Generally, a small period of time is reserved at the beginning of the policy event time window to be unaffected by the policy, as anomaly detection algorithms typically rely on continuous observation of normal conditions to identify the onset of sudden or abnormal behavior. If price anomalies occur at the beginning of the event window and the system does not reserve this transition period for behavioral comparison, it will be difficult to determine whether the current fluctuations constitute abnormal fluctuations. Therefore, establishing such an "observation lead time" free of policy disturbances ensures that the algorithm has sufficient reference context within the event window, thereby improving its ability to identify and respond promptly to unexpected price fluctuations. For example, if, through analysis of policy events over the past few years, market price fluctuations typically begin 30 days before a policy release, the policy event time window can be set to 40 days before the policy release.

[0075] Once the time window is determined, the next step is to collect daily average quote data for the target drug within that time period. This data can be obtained from a centralized pharmaceutical procurement platform, a supplier quotation system, or a market transaction database. In some embodiments, to ensure data integrity and accuracy, a preprocessing mechanism can be incorporated into the collection process.

[0076] For example, after collecting quotes from multiple suppliers every day, the average value of the day is calculated as P t , which is the daily average quote. If quote data for a particular day is missing, it can be filled in through interpolation (for example, using the average of the previous and next day's quotes). If a supplier's quote is found to be significantly out of the normal range (for example, exceeding 5 standard deviations from the mean), it will be considered an outlier and removed.

[0077] In some embodiments, data collection can occur daily. However, for drugs with significant price fluctuations, data collection can be increased to multiple times daily (e.g., once in the morning, afternoon, and evening) to more accurately capture price trends. Furthermore, the system can record each supplier's quote data to facilitate subsequent calculation of the supplier's synchronization response index.

[0078] like Figure 2 As shown in Figure 2, after collecting the daily average quote data, it is necessary to normalize the data to eliminate the differences in absolute price levels and generate standardized series data that can reflect the degree of volatility. The formula for normalization is:

[0079]

[0080] Among them, Z t is the standardized price data, P t is the average daily quote at time t, μ is the mean of the price during the stable period, and σ is the standard deviation during the stable period.

[0081] The significance of this formula is that it compares the daily price with the benchmark of the stable period and converts it into a dimensionless number, thereby highlighting the relative volatility of the price. For example, if the price on a certain day is two standard deviations above the stable period mean, Z t =2, indicating that the price fluctuation is more significant.

[0082] A stable period refers to a period of time when prices are relatively stable and are not affected by policy events. Typically, a longer period before the policy is released can be selected, such as the previous 12 months (i.e., 365 days before the policy is released). During this period, the present invention calculates the mean μ and standard deviation σ of all daily average quotes. To ensure the representativeness of the stable period, sub-periods affected by other policies or abnormal events can be excluded. For example, if another policy adjustment occurred in the past year, the data for that period can be excluded, and only the truly stable monthly data can be retained.

[0083] like Figure 3 As shown in Figure 2, after the standardization process is completed, the generated standardized sequence data is input into the anomaly detection algorithm to identify price anomalies in the policy event time window.

[0084] In the embodiment of the present invention, a cumulative sum detection algorithm (CUSUM) can be used. The cumulative sum detection algorithm is based on L t =max(0,L t-1 +Z t -k) Update the cumulative amount L t , if L t If the set control limit h is exceeded, t is determined to be an abnormal point; the formula means: the daily standardized fluctuation value Z t The difference between the two values ​​and a pre-set baseline deviation constant k is accumulated. If the accumulated result is less than 0, it is reset to 0, which means that the algorithm only focuses on the behavior of "continuous upward (or downward) deviation". In other words, when the quotes for several consecutive days deviate abnormally in a certain direction, the accumulated value L t It will gradually increase until it exceeds a preset threshold h, and the system will determine that this time point is a significant abnormal fluctuation.

[0085] The baseline shift constant k serves to define the "normal range of fluctuations," distinguishing minor normal changes from persistent abnormal shifts. A common setting is to set k as a fraction or multiple of the standard deviation σ, for example, k = 0.5σ or k = δ. In some embodiments, the threshold h can be set based on empirical standard deviations, for example, h = 4-5·σ, or four to five times the standard deviation σ.

[0086] In other embodiments, an exponentially weighted moving average (EWMA) algorithm can also be used. The EWMA algorithm smoothes data through exponentially weighted averaging and is suitable for detecting price trends. It assigns higher weights to recent data and can effectively capture sustained upward or downward price trends. Since the cumulative sum detection algorithm and the exponentially weighted moving average algorithm are existing technologies for detecting standardized sequence data, they will not be discussed in detail here.

[0087] like Figure 4 As shown, for each abnormal time point, its abnormality score needs to be calculated to comprehensively evaluate the severity of the abnormality. The formula for the abnormality score is:

[0088]

[0089] The above formula combines three key factors. The following explains their meaning and implementation one by one:

[0090] Item 1 This term represents the lead time between the anomaly time point and the policy release date, where T is the policy release date and t is the date of the anomaly time point. The logic behind this term is that the closer the anomaly occurs to the policy release date, the higher its correlation with the policy, and thus its importance, as it may reflect strong market anticipation or early reaction to the policy.

[0091] The second term R t is the supplier's synchronous response index, calculated as:

[0092]

[0093] Where N2(t) is the number of all suppliers who submitted quotations at time t, and N1(t) is the number of suppliers whose quotations at the abnormal time point t have an absolute value that exceeds the threshold δ compared with the mean μ.

[0094] In some embodiments, the threshold δ is calculated as follows:

[0095]

[0096] Here, α is an adjustment coefficient, ranging from 1.5 to 3, indicating that fluctuations exceeding 1.5 to 3 standard deviations are abnormal. This index reflects the consistency of suppliers adjusting their quotes at the same time, which may indicate market linkage or collusion.

[0097] Item 3 This is the absolute value of the normalized price divided by the standard deviation, reflecting the magnitude of price fluctuations. This term is set to quantify the intensity of anomalies; greater fluctuations are associated with higher anomaly scores.

[0098] In the formula, w1, w2, and w3 are weighting coefficients used to balance the importance of the three factors. These coefficients typically range from 0 to 1, satisfying the requirement w1 + w2 + w3 = 1. Specific values ​​can be determined based on specific circumstances. For example, by analyzing the correlation between anomaly scores of past policy events and actual market reactions, w1 = 0.3, w2 = 0.4, and w3 = 0.3. If the warning effect of supplier synchronization is desired, the value of w2 can be appropriately increased, for example, to 0.5.

[0099] After calculating the anomaly score, the system compares it with the preset threshold. t If the threshold is exceeded, an early warning message of abnormal quotation for the target drug will be output.

[0100] In some embodiments, S t The threshold can be set through historical sample statistics, that is, in a normal period without policy event intervention, the quotation data of drugs of the same or similar categories are collected, and the typical distribution range of abnormal scores under these circumstances is calculated. Based on the historical distribution, the system extracts the higher percentile (such as the 90th percentile) as the benchmark threshold for risk judgment. When the abnormal score of the target drug within the policy event window exceeds the historical threshold, it is considered that there is an abnormal early quotation behavior, and an early warning prompt is triggered. This method does not need to rely on manually set fixed values, and can dynamically adapt to different drug categories and market volatility characteristics, thereby improving the sensitivity and rationality of detection. The early warning information can include abnormal time points, abnormal scores, price fluctuation details, and synchronous response index, which are sent to regulatory authorities, procurement platforms, medical structures, etc. through a visual interface or report form.

[0101] A second aspect of the present invention also discloses an apparatus corresponding to the above method, comprising:

[0102] An event window setting unit, configured to set a policy event time window for a target drug, wherein the policy event time window includes a period of time before the policy release date;

[0103] a quotation data collection unit, configured to collect daily average quotation data of target drugs within the policy event time window;

[0104] a normalization processing unit, configured to perform normalization processing on the daily average quotation data based on the mean and standard deviation of a period of price stability before the policy event time window, to generate normalized series data reflecting the degree of volatility;

[0105] an anomaly detection unit, configured to input the standardized sequence data into an anomaly detection algorithm, identify price anomaly point data appearing in the policy event time window, and form a corresponding anomaly time point set;

[0106] An indicator calculation unit is used to calculate, for each price anomaly data point, the lead time from the corresponding anomaly time point date to the policy release date, calculate the synchronous response index of multiple suppliers at the same anomaly time point date, and the price in the standardized series data at the same anomaly time point date;

[0107] An early warning output unit is used to construct an abnormality score based on the advance response time, the synchronous response index and the price in the standardized sequence data, and when the abnormality score exceeds a preset threshold, output early abnormal quotation warning information for the target drug.

[0108] As can be seen above, this invention, by introducing policy event time windows, standardized processing, and multi-dimensional anomaly scoring, can effectively identify policy-driven price anomalies and reveal market-linked behaviors. Compared to traditional monitoring methods, it significantly improves timeliness, accuracy, and adjustability, making it suitable for the drug price management needs of regulatory agencies, medical institutions, and centralized procurement platforms.

[0109] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A data monitoring method for pharmaceutical collection channels, characterized in that: The steps include: Setting a policy event time window for the target drug, where the policy event time window includes a period of time before the policy issuance date; Collecting daily average quotation data of target drugs within the policy event time window; Normalizing the daily average quotation data based on the mean and standard deviation of a period of price stability before the policy event time window to generate standardized series data reflecting the degree of volatility; Inputting the standardized sequence data into an anomaly detection algorithm to identify price anomaly point data appearing in the policy event time window and forming a corresponding anomaly time point set; For each price anomaly point data, calculate the lead time from the corresponding abnormal time point date to the policy release date, calculate the synchronous response index of multiple suppliers at the same abnormal time point date, and the price in the standardized series data at the same abnormal time point date; constructing an anomaly score based on the lead response time, the synchronous response index, and prices in the standardized sequence data; When the abnormality score exceeds a preset threshold, early abnormal quotation warning information for the target drug is output.

2. The data monitoring method for pharmaceutical collection channels according to claim 1, characterized in that: The corresponding formula for the standardization process is: where Z t is the standardized sequence data that changes with time t after standardization, P t is the average daily quote, μ is the mean of the price during the price stability period, and σ is the standard deviation of the price during the price stability period.

3. The data monitoring method for pharmaceutical collection channels according to claim 1, characterized in that: The anomaly detection algorithm is a cumulative sum detection algorithm CUSUM.

4. The data monitoring method for pharmaceutical collection channels according to claim 2, characterized in that: The formula for the anomaly score is: in: S t is the anomaly score corresponding to the abnormal time point date t; T is the policy issuance date; R t is the supplier's synchronous response index at the abnormal time point date t; Z t is the standardized price; σ is the standard deviation of prices during the period of price stability; w1, w2, w3 are the corresponding weight coefficients.

5. The data monitoring method for pharmaceutical collection channels according to claim 4, characterized in that: R t The calculation formula is: Where N2(t) is the number of all suppliers who submitted quotations at the abnormal time point date t; N1(t) is the number of suppliers whose absolute value of the difference between their quotations and the mean μ at the abnormal time point t exceeds the threshold δ.

6. The data monitoring method for pharmaceutical collection channels according to claim 5, characterized in that: The threshold value is calculated as follows: Among them, μ is the mean of the price during the price stability period, σ is the standard deviation of the price during the price stability period, and α is the adjustment coefficient.

7. A data monitoring device for medical collection channels, characterized in that: include: An event window setting unit, configured to set a policy event time window for a target drug, wherein the policy event time window includes a period of time before the policy release date; a quotation data collection unit, configured to collect daily average quotation data of target drugs within the policy event time window; a normalization processing unit, configured to perform normalization processing on the daily average quotation data based on the mean and standard deviation of a period of price stability before the policy event time window, to generate normalized series data reflecting the degree of volatility; an anomaly detection unit, configured to input the standardized sequence data into an anomaly detection algorithm, identify price anomaly point data appearing in the policy event time window, and form a corresponding anomaly time point set; An indicator calculation unit is used to calculate, for each price anomaly data point, the lead time from the corresponding anomaly time point date to the policy release date, calculate the synchronous response index of multiple suppliers at the same anomaly time point date, and the price in the standardized series data at the same anomaly time point date; An early warning output unit is used to construct an abnormality score based on the advance response time, the synchronous response index and the price in the standardized sequence data, and when the abnormality score exceeds a preset threshold, output early abnormal quotation warning information for the target drug.

8. The data monitoring device for pharmaceutical collection channels according to claim 7, characterized in that: The corresponding formula for the standardization process is: where Z t is the standardized sequence data that changes with time t after standardization, P t is the average daily quote, μ is the mean of the price during the price stability period, and σ is the standard deviation of the price during the price stability period.

9. The data monitoring device for pharmaceutical collection channels according to claim 7, characterized in that: The formula for the anomaly score is: in: S t is the anomaly score corresponding to the abnormal time point date t; T is the policy issuance date; R t is the supplier's synchronous response index at the abnormal time point date t; Z t is the standardized price; σ is the standard deviation of prices during the period of price stability; w1, w2, w3 are the corresponding weight coefficients.

10. The data monitoring device for pharmaceutical collection channels according to claim 7, characterized in that: R t The calculation formula is: Where N2(t) is the number of all suppliers who submitted quotations at the abnormal time point date t; N1(t) is the number of suppliers whose absolute value of the difference between their quotations and the mean μ at the abnormal time point t exceeds the threshold δ.

Images