An intelligent verification method and system for a vaccination process

Abstract

This invention relates to the field of medical information management technology, specifically to an intelligent verification method and system for vaccine administration processes. In this invention, isolated operation logs during the vaccination process are reconstructed into a complete sequence of recipient vaccination behaviors based on their inherent temporal correlation and the homology of the executing entities. A Hidden Markov Model is then used to perform in-depth analysis of this sequence, inferring the underlying vaccination stage states that conform to clinical logic from the observed specific behaviors. This elevates verification from static data checking to a dynamic process review. Furthermore, the inferred actual stage sequence is rigorously compared with the standardized process in three dimensions: stage completeness, sequential accuracy, and dose interval. Innovatively, a probability confidence level is calculated for any identified process deviations, achieving a quantitative assessment of abnormal behavior.

Description

Technical Field

[0001] This invention relates to the field of medical information management technology, and in particular to an intelligent verification method and system for the vaccination process. Background Technology

[0002] The field of medical information management technology involves the collection, storage, transmission, processing and application of medical data. Its core aspects include the management of basic patient information, the construction of electronic medical record systems, the standardization of clinical data, and the implementation of medical resource scheduling and information sharing mechanisms.

[0003] Among them, the intelligent verification method used in the vaccination process refers to the technical approach of systematically verifying the recipient's identity information, vaccination records, vaccine batches, and vaccination schedule during the vaccination process. This involves reading the recipient's ID number, health record number, or electronic vaccination code through the vaccination information system and comparing it with the vaccination schedule and inventory data stored in the vaccine management database to confirm the identity of the person receiving the vaccination and the matching relationship with the corresponding vaccine.

[0004] In the verification of the vaccination process, the existing technology mainly relies on the real-time comparison of static data such as recipient identity information, vaccine batches, and vaccination plans. This verification method is essentially an access control check, which can only confirm whether the matching relationship between the subject and the item is correct at a specific point in time before vaccination. The system lacks the ability to continuously perceive and logically analyze the entire actual operation process. Therefore, when vaccination personnel make dynamic behavioral deviations such as omitting key steps, reversing the order of operations, or failing to follow the correct dose intervals, the existing technology cannot effectively identify these procedural errors because it cannot interpret discrete operation logs as a complete process with internal logic. Ultimately, this leads to the omission of potential violations and safety hazards, and the inability to guarantee the standardization of the entire vaccination process. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent verification method and system for the vaccination process.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent verification method for a vaccination process, comprising the following steps:

[0007] S1: Obtain information about the vaccine recipient and the vaccination implementer during the vaccination process, match the information with the preset vaccine management database, determine the same-origin vaccination behavior, and construct the vaccine recipient's vaccination behavior sequence;

[0008] S2: Input the recipient's vaccination behavior sequence into a hidden Markov model, calculate the posterior probability of each vaccination behavior under a preset vaccination stage state set, arrange the vaccination stage states with the highest posterior probability in time, and generate a vaccination stage sequence.

[0009] S3: Obtain the recipient's dose registration number and vaccine type code record, establish a dose stage correspondence table according to the vaccination time sequence, verify the order of each vaccination stage status in the vaccination stage sequence according to the dose stage correspondence table, and obtain the vaccination consistency record;

[0010] S4: Based on the vaccination consistency record, filter abnormal vaccination cases, mark the abnormal vaccination type, and construct an abnormal vaccination operation list;

[0011] S5: Determine the probability confidence level of each type of abnormal vaccination operation in the abnormal vaccination operation list, allocate early warning instructions according to the confidence level interval, and obtain the vaccination process verification results.

[0012] As a further aspect of the present invention, the recipient vaccination behavior sequence includes a vaccination behavior identifier, a behavior time index, and a behavior type label; the vaccination stage sequence includes a stage status name, a stage time index, and a stage corresponding probability value; the vaccination consistency record includes a stage sequence structure, a dose correspondence, and a time connection status; the abnormal vaccination operation list includes an abnormality type category, an abnormality occurrence location, and an abnormality association number; and the vaccination process verification result includes an abnormality probability confidence interval, a warning instruction level, and a recipient corresponding identifier.

[0013] As a further aspect of the present invention, the step of obtaining the recipient's vaccination behavior sequence specifically includes:

[0014] S111: Obtain information about the vaccine recipient and the vaccine executor during the vaccination process, including the recipient's unique identifier, device identification number, operator identification number, vaccine operation type and vaccine operation time. Based on the recipient's unique identifier, filter all corresponding vaccination log records in the preset vaccine management database, determine the consistency between the device identification number and the operator identification number, and establish the same source vaccination behavior determination result.

[0015] S112: Based on the determination result of the same vaccination behavior, all vaccination log records of the same recipient are aggregated, and the vaccination operation type corresponding to each operation event is matched with the preset vaccination event type, including information registration, vaccine preparation, vaccine injection and result confirmation, to obtain the aggregated vaccination event type mapping set.

[0016] S113: Based on the aggregated vaccination event type mapping set, all vaccination operations for the same recipient are arranged in ascending order of vaccination operation time, and each operation event is treated as a vaccination behavior record to generate a recipient vaccination behavior sequence.

[0017] As a further aspect of the present invention, the step of obtaining the vaccination stage sequence specifically includes:

[0018] S211: Define the vaccination operation type of each vaccination behavior record in the vaccination behavior sequence of the recipient as an observation variable, calculate the time difference between adjacent vaccination operation times as a time feature, and input the observation variable and time feature into a hidden Markov model to establish a vaccination behavior observation feature set.

[0019] S212: Based on the observation feature set of the vaccination behavior and the initial state distribution and state transition probability of the hidden Markov model, for the preset vaccination stage state set, including the registration stage, preparation stage, injection stage and confirmation stage, calculate the posterior probability of each observation behavior in each vaccination stage state, and obtain the posterior probability set of each stage state.

[0020] S213: For each stage state posterior probability group, the maximum a posteriori decision rule is applied to each vaccination action to select the vaccination stage state with the largest posterior probability value and arrange them according to the vaccination operation time sequence to generate a vaccination stage sequence.

[0021] As a further aspect of the present invention, the process of combining observed variables and time features and inputting them into a hidden Markov model to establish a set of observed features of vaccination behavior is as follows:

[0022] Multiple time thresholds are preset, and the time thresholds are determined based on the statistical distribution of historical vaccination operation time data;

[0023] The time features are quantized into intervals based on time thresholds and mapped to discrete time interval labels.

[0024] The vaccination operation type of each vaccination behavior record in the vaccination behavior sequence is taken as a discrete observation category, and a binary observation group is formed by the time interval label mapped by the time feature of each vaccination behavior record. The observation probability of the Hidden Markov Model is calculated based on the binary observation group.

[0025] All binary observation groups are arranged in chronological order of vaccination operation to form an observation sequence, which is the observation feature set of vaccination behavior.

[0026] As a further aspect of the present invention, the step of obtaining the vaccination consistency record specifically includes:

[0027] S311: Obtain the dose registration number and vaccine type code record of the recipient, call the vaccination stage sequence, match and associate the dose registration number and vaccine type code record with the status of each vaccination stage in the sequence, and establish a dose stage correspondence table;

[0028] S312: For each dose in the dose stage correspondence table, the vaccination stage state sequence associated with each dose is compared with the flow of the vaccination stage state set item by item to check the completeness of the vaccination stage state sequence and the accuracy of the order of arrangement, and to obtain the stage order logic check result.

[0029] S313: Based on the dose-stage correspondence table, calculate the time difference of vaccination operation between adjacent doses, compare the time difference of vaccination operation with the planned vaccination interval in the corresponding vaccine type code record to obtain the dose-stage time interval comparison result, integrate it with the stage sequence logic test result, and generate a vaccination consistency record.

[0030] As a further aspect of the present invention, the step of obtaining the abnormal inoculation operation list specifically includes:

[0031] S411: Retrieve the comparison results reflected in the stage sequence logic test results within the vaccination consistency record, filter records with incomplete or incorrect stage sequence results, and generate an abnormal vaccination situation screening set by recording the time interval between doses that is less than the planned vaccination interval.

[0032] S412: For each record in the abnormal vaccination situation screening set, the abnormal content is classified into three preset abnormal vaccination types: missing stage order, disordered stage order, or abnormal dose interval. Type labels are added to the records to establish an abnormal vaccination type label library.

[0033] S413: Based on the abnormal vaccination type labeling library, extract the unique identifier of the recipient, the device identification number, and the operator's identifier corresponding to each record, and group and arrange them according to the labeled abnormal vaccination types to construct an abnormal vaccination operation list.

[0034] As a further aspect of the present invention, the steps for obtaining the verification results of the vaccination process are specifically as follows:

[0035] S511: For each abnormal vaccination type in the abnormal vaccination operation list, extract the corresponding posterior probability from the posterior probability group of each stage state, and perform aggregate operation on the posterior probabilities of all abnormal operations under each abnormal vaccination type to obtain the probability confidence of abnormal vaccination operation.

[0036] S512: Set up multi-level confidence intervals, compare the probability confidence of each type of abnormal vaccination operation with the confidence interval, and assign one of the three types of early warning instructions (time out of bounds, stage disorder, and dose skipping) according to the range of the interval, and obtain a graded early warning instruction set;

[0037] S513: Invoke the tiered early warning instruction set, extract the recipient's unique identifier from the abnormal vaccination operation list, obtain the dose registration number from the dose stage correspondence table, integrate the early warning instruction with the recipient's unique identifier and dose registration number, and generate the vaccination process verification result.

[0038] A smart verification system for a vaccination process, the smart verification system for a vaccination process being used to execute the aforementioned smart verification method for a vaccination process, the system comprising:

[0039] The vaccination behavior data acquisition module acquires information about vaccine recipients and vaccination implementers during the vaccination process, matches the information with a preset vaccination management database, determines homologous vaccination behaviors, and constructs a vaccine recipient vaccination behavior sequence.

[0040] The vaccination stage inference module inputs the recipient's vaccination behavior sequence into a hidden Markov model, calculates the posterior probability of each vaccination behavior under a preset vaccination stage state set, arranges each vaccination stage state with the highest posterior probability in time, and generates a vaccination stage sequence.

[0041] The dose-stage consistency verification module obtains the recipient's dose registration number and vaccine type code record, establishes a dose-stage correspondence table according to the vaccination time sequence, verifies the order of each vaccination stage status in the vaccination stage sequence based on the dose-stage correspondence table, and obtains the vaccination consistency record.

[0042] The abnormal vaccination identification module filters abnormal vaccination cases based on the vaccination consistency record, marks the abnormal vaccination type, and constructs an abnormal vaccination operation list.

[0043] The process verification and archiving module determines the probability confidence level of each type of abnormal vaccination operation in the abnormal vaccination operation list, allocates early warning instructions according to the confidence level interval, and obtains the vaccination process verification results.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0045] In this invention, isolated operation logs during the vaccination process are reconstructed into a sequence of recipient vaccination behaviors that fully reflect the operation process based on their inherent temporal correlation and the homology of the executing entities. A Hidden Markov Model is then used to conduct in-depth analysis of this sequence, inferring the underlying vaccination stage status that conforms to clinical logic from the observed specific behaviors. This elevates the verification from static data checking to dynamic process review. Furthermore, the inferred actual stage sequence is rigorously compared with the standardized process in three dimensions: stage completeness, sequential accuracy, and dose interval. Innovatively, a probability confidence level is calculated for any identified process deviations, achieving a quantitative assessment of abnormal behavior. Finally, based on the confidence level, corresponding warning instructions are triggered, accurately identifying and pointing out procedural errors that cannot be detected by traditional methods. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0047] Figure 2 This is a flowchart of step S1 of the present invention;

[0048] Figure 3 This is a flowchart of step S2 of the present invention;

[0049] Figure 4 This is a flowchart of step S3 of the present invention;

[0050] Figure 5 This is a flowchart of step S4 of the present invention;

[0051] Figure 6 This is a flowchart of step S5 of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0053] Please see Figure 1 This invention provides a technical solution: an intelligent verification method for the vaccination process, comprising the following steps:

[0054] S1: Obtain information about the vaccine recipient and the vaccination implementer during the vaccination process, match the information with the preset vaccine management database, determine the same-origin vaccination behavior, and construct the vaccine recipient's vaccination behavior sequence;

[0055] S2: Input the recipient's vaccination behavior sequence into the Hidden Markov Model, calculate the posterior probability of each vaccination behavior under the preset vaccination stage state set, arrange the vaccination stage states with the highest posterior probability in time, and generate the vaccination stage sequence.

[0056] S3: Obtain the recipient's dose registration number and vaccine type code record, establish a dose stage correspondence table according to the vaccination time sequence, verify the order of each vaccination stage status in the vaccination stage sequence according to the dose stage correspondence table, and obtain the vaccination consistency record;

[0057] S4: Filter abnormal vaccination cases based on vaccination consistency records, mark the abnormal vaccination types, and construct an abnormal vaccination operation list;

[0058] S5: Determine the probability confidence level of each type of abnormal vaccination operation in the abnormal vaccination operation list, allocate early warning instructions according to the confidence level interval, and obtain the verification results of the vaccination process.

[0059] The recipient's vaccination behavior sequence includes vaccination behavior identifier, behavior time index, and behavior type label; the vaccination stage sequence includes stage status name, stage time index, and corresponding probability value; the vaccination consistency record includes stage sequence structure, dose correspondence, and time connection status; the abnormal vaccination operation list includes abnormal type category, abnormal occurrence location, and abnormal association number; and the vaccination process verification results include abnormal probability confidence interval, warning instruction level, and recipient's corresponding identifier.

[0060] Please see Figure 2 The specific steps for obtaining the recipient's vaccination behavior sequence are as follows:

[0061] S111: Obtain information about the vaccine recipient and the vaccine executor during the vaccination process, including the recipient's unique identifier, device identification number, operator identification number, vaccine operation type and vaccine operation time. Based on the recipient's unique identifier, filter all corresponding vaccination log records in the preset vaccine management database, determine the consistency between the device identification number and the operator identification number, and establish the same source vaccination behavior determination result.

[0062] The system extracts all vaccination log records associated with a specific recipient's unique identifier from the vaccination management database. For example, for a child with the unique identifier "PID20251106001", the system retrieves all log entries since their first vaccination. Each log record includes the device identification number, operator identification, vaccination operation type, and vaccination operation time. Next, a homology analysis is performed on all selected vaccination log records. This homology analysis is not a simple information comparison but a clustering judgment based on the continuity of operations. Specifically, the system uses the vaccination operation time as a basis, considering consecutive records with timestamp differences within a preset "session interval threshold" as a potential vaccination session. This "session interval threshold" is set as follows: First, the system statistically analyzes the time intervals between adjacent operation steps within the same vaccination session in historical compliant vaccination data, forming a time interval dataset. Then, the average and maximum values ​​of this dataset are calculated. Ultimately, the threshold is set to the smaller of a multiple of the average (e.g., 3 times) and a proportion of the maximum value (e.g., 70%). This method aims to ensure that the threshold is both tolerant of fluctuations in the time consumption of routine operations and sensitive enough to separate independent vaccination services, thereby guaranteeing the complete capture of a single vaccination behavior. Within each identified potential vaccination session, the consistency of all recorded device IDs and operator identities is checked. For example, in a vaccination session “PID20251106001”, five operations were recorded. The operator identity for the first four records was “UID034”, and the device ID was “DEV551”. However, the operator identity for the fifth record changed to “UID035”, while the device ID remained “DEV551”. In this case, the system classifies the first four records as a group of vaccination behaviors from the same source and the fifth record as another group. This judgment is based on the fact that within the same vaccination session, a change in the operator or device signifies a change in the subject performing the vaccination, and therefore it is necessary to separate them into different behavioral sources. If all records within a session retain the same device identifier and operator identity, then all records within the entire session constitute a set of homologous vaccination behaviors. Ultimately, the system assigns a homologous determination tag to each vaccination log record. This tag uniquely identifies the executing entity and device combination to which it belongs, thereby establishing the homologous vaccination behavior determination result.

[0063] S112: Based on the results of the same vaccination behavior determination, all vaccination log records for the same recipient are aggregated, and the vaccination operation type corresponding to each operation event is matched with the preset vaccination event type, including information registration, vaccine preparation, vaccine injection and result confirmation, to obtain the aggregated vaccination event type mapping set.

[0064] Aggregation is performed on all vaccination log records marked as originating from the same source and belonging to the same recipient. Taking a set of homologous vaccination log records generated by recipient "PID20251106001" in a single vaccination service as an example, this set of records includes original vaccination operation types such as "scanning recipient identification code," "scanning vaccine supervision code," "performing intramuscular injection," and "confirming vaccination completion." The next step is to match and map these original vaccination operation types to a pre-defined set of vaccination event types. This pre-defined set of vaccination event types is fixed and specifically includes four types: information registration, vaccine preparation, vaccine injection, and result confirmation. The mapping process is based on a well-defined mapping rule base, which defines the vaccination event type to which each original operation type should belong. This rule base is established based on semantic analysis and operational decomposition of the National Health Commission's "Standard Operating Procedures for Vaccination" (SOP), mapping each standard action defined in the official procedure to a one-to-one correspondence with the original operation types that can be recorded in the system. For example, operations such as "scanning the recipient's ID code" and "reading ID card information" are mapped to the "information registration" event type; operations such as "scanning the vaccine regulatory code," "verifying the vaccine batch number," and "extracting the medication" are mapped to the "vaccine preparation" event type; operations such as "performing an intramuscular injection" and "subcutaneous injection" are mapped to the "vaccine injection" event type; and operations such as "confirming vaccination completion" and "printing the vaccination certificate" are mapped to the "result confirmation" event type. When processing the log records of recipient "PID20251106001," the system reads each vaccination operation type and searches for the corresponding vaccination event type in the rule base. For example, a record with the operation type "scanning the vaccine regulatory code" is matched, and its corresponding vaccination event type is determined to be "vaccine preparation." By performing this mapping process on all log records of this recipient under the same source behavior, the originally scattered operation records are transformed into a set composed of standardized vaccination event types, resulting in an aggregated vaccination event type mapping set.

[0065] S113: Based on the aggregated vaccination event type mapping set, all vaccination operations for the same recipient are arranged in ascending order of vaccination operation time, and each operation event is treated as a vaccination behavior record to generate a recipient vaccination behavior sequence.

[0066] Each mapped vaccination event in the set is re-associated with the vaccination operation time in its original record. Then, all these time-associated vaccination events are sorted in ascending order strictly according to the chronological order of the vaccination operation time. For example, for recipient "PID20251106001", its mapped vaccination event type set might include "Information Registration", "Vaccine Preparation", "Vaccine Injection", and "Result Confirmation", with corresponding original operation timestamps of "2025-11-06-09:30:15", "2025-11-06-09:32:40", "2025-11-06-09:35:10", and "2025-11-06-09:50:20" respectively. The sorting process involves organizing these events in ascending order of timestamps to form an ordered list. After sorting, each operation event in this ordered list (i.e., the mapped vaccination event type and its corresponding timestamp) is defined as an independent vaccination behavior record. Each vaccination record precisely describes the standardized vaccination event that occurred at a specific point in time. For example, the first vaccination record is "09:30:15, Information Registration," the second is "09:32:40, Vaccine Preparation," and so on. By converting all the ordered operational events into vaccination records one by one while maintaining their chronological order, an orderly whole composed of multiple vaccination records is ultimately constructed that fully reflects the process of a single vaccination service, generating a sequence of vaccination behaviors for the recipient.

[0067] Please see Figure 3 The specific steps for obtaining the vaccination phase sequence are as follows:

[0068] S211: Define the vaccination operation type of each vaccination behavior record in the vaccination behavior sequence as the observation variable, calculate the time difference between adjacent vaccination operation times as the time feature, and input the observation variable and the time feature into the hidden Markov model to establish the vaccination behavior observation feature set.

[0069] The process of combining observed variables with time features and inputting them into a hidden Markov model to establish a set of observed features of vaccination behavior is as follows:

[0070] Multiple time thresholds are preset, and the time thresholds are determined based on the statistical distribution of historical vaccination operation time data;

[0071] The time features are quantized into intervals based on time thresholds and mapped to discrete time interval labels.

[0072] The vaccination operation type of each vaccination behavior record in the vaccination behavior sequence is taken as a discrete observation category, and a binary observation group is formed by the time interval label mapped by the time feature of each vaccination behavior record. The observation probability of the Hidden Markov Model is calculated based on the binary observation group.

[0073] All binary observation groups are arranged in chronological order of vaccination operation to form an observation sequence, which is the observation feature set of vaccination behavior.

[0074] The vaccination operation type included in each vaccination behavior record in the recipient's vaccination behavior sequence is directly defined as the observed variable in the Hidden Markov Model. Simultaneously, for each pair of temporally adjacent vaccination behavior records in the sequence, the difference between the vaccination operation time of the later record and the vaccination operation time of the earlier record is calculated; this difference is defined as the time feature. For example, if the operation time for "Vaccine Preparation" is 09:32:40 and the operation time for "Information Registration" is 09:30:15, then the time feature value between the two is 145 seconds. Next, the calculated time feature is discretized. This processing relies on a set of preset multiple time thresholds. These time thresholds are determined as follows: First, a large amount of historical compliant vaccination operation time data is collected, and for each pair of adjacent operation types (e.g., "Information Registration" to "Vaccine Preparation"), the distribution of their time differences is statistically analyzed. Then, the statistical quantiles of each set of time difference data are calculated, such as the 25th quantile, 50th quantile (median), and 75th quantile. These three quantiles are set as the time thresholds for this operation type pair, used to divide continuous time differences into different discrete intervals. To verify the rationality of the thresholds, a ten-fold cross-validation method is used. The historical dataset is divided into ten parts, and nine parts are used alternately as the training set to calculate the quantile thresholds, with the remaining part used as the test set. On the test set, the generated thresholds are used to quantize time features and input into the subsequent Hidden Markov Model for stage identification. The accuracy of the model identification is recorded. After repeating this process ten times, the average of the ten accuracy rates is calculated. If the average accuracy is higher than the preset baseline accuracy, it proves that the threshold set based on quantiles is stable and effective. The "baseline accuracy" here is set based on the minimum reliability requirements for business verification of the vaccination process, and is usually set to a value close to but slightly lower than the theoretical optimal model performance (e.g., 95%) to ensure the effectiveness of the method in practical applications. Using the determined time thresholds, continuous time feature values ​​are mapped to discrete time interval labels. For example, the aforementioned time feature of 145 seconds is mapped to the time interval label "normal" because it falls between 90 seconds (25th percentile) and 150 seconds (50th percentile). If it's less than 90 seconds, it's "too fast"; between 150 and 240 seconds (75th percentile), it's "slow"; and greater than 240 seconds, it's "too slow." Subsequently, the vaccination operation type (as a discrete observation category) of each vaccination behavior record in the recipient's vaccination behavior sequence is combined with the time interval label mapped by its time feature and the previous behavior record to form a binary observation group. For example, for the behavior record "vaccine preparation," its observation variable is "vaccine preparation," and the time interval label is "normal," forming a binary observation group (vaccine preparation, normal). The observation probability of the Hidden Markov Model is calculated based on this binary observation group.Finally, all the binary observation sets generated from the operation events are arranged strictly according to the original chronological order of the vaccination operation to form an observation sequence. This observation sequence serves as the set of vaccination behavior observation features as input to the Hidden Markov Model.

[0075] S212: Based on the observation feature set of vaccination behavior and the initial state distribution and state transition probability of the hidden Markov model, for the preset vaccination stage state set, including the registration stage, preparation stage, injection stage and confirmation stage, calculate the posterior probability of each observation behavior in each vaccination stage state, and obtain the posterior probability set of each stage state.

[0076] The structure of the Hidden Markov Model (HMM) used is clearly defined. This model includes: 1) a set of unobservable hidden states, i.e., four pre-defined vaccination stages (registration, preparation, injection, and confirmation); 2) a set of observed variables, i.e., the binary observation set generated in S211 (vaccination operation type, time interval label); 3) an initial state distribution probability, defining the probability of being in each hidden state at the start of the process; 4) a state transition probability matrix, defining the probability of transitioning from one hidden state to another; and 5) an observation probability matrix (or emission probability matrix), defining the probability of generating a specific observed variable in a given hidden state. The initial state distribution, state transition probabilities, and observation probabilities of the model are all obtained through training and learning using the Baum-Welch algorithm on a large amount of historical compliant vaccination sequence data. The model execution process is as follows: Based on the vaccination behavior observation feature set and the trained HMM parameters, for the pre-defined vaccination stage state set, the forward-backward algorithm is used to calculate the posterior probability of each observed behavior in each vaccination stage state. Let the observation sequence be... , which includes One observation. Given a complete sequence of observations. and model parameter set Under these conditions, the system at time step In the first Status at each stage of vaccination The posterior probability is calculated using the following formula: ;in, Represents the system at time step The hidden state, Representing the A preset vaccination stage state, for example For the registration phase, For the preparation stage, Represents a complete set of observational features of vaccination behavior, such as , This represents the complete parameter set of the Hidden Markov Model, including the state transition probability matrix, observation probability matrix, and initial state distribution vector. These parameters are all derived from historical compliant vaccination data. This represents the total number of preset vaccination stage states, which is 4 in this embodiment. It is a forward variable, representing the time step. Partial sequence observed And the system is in state The joint probability, It is a backward variable, representing the time step. The system is in a state Under these conditions, subsequent sequences were observed. The conditional probability, It is an index variable used to traverse all possible states in the summation calculation, ranging from 1 to... denominator It is a normalization factor that ensures that at any time step... The sum of the posterior probabilities of all states is 1.

[0077] Suppose we need to calculate in At that time, the system is in state The posterior probability (during the preparation phase), i.e. 1. Setting the scene and parameters: Observation sequence These correspond to (Registration, Normal), (Preparation, Normal), (Injection, Normal), and (Confirmation, Normal), respectively. (Status Set) These correspond to the registration, preparation, injection, and confirmation stages, respectively. Model parameters (Originally trained from historical data): Initial state distribution State transition matrix Observation probability matrix (Excerpt) The probability of observing a non-corresponding operation in other states is set to a small value, such as 0.05.

[0078] 2. Calculate the forward variables First calculate :

[0079] , , , ;

[0080] Next calculation :

[0081] , , , ;

[0082] 3. Calculate the backward variables First, set For all .

[0083] calculate :

[0084] , , , ;

[0085] calculate :

[0086] (because ), (because ),because and The calculation result is 0, and when summing the denominators, the corresponding product terms are... and The value must be 0, therefore no calculation is needed here. and .

[0087] 4. Calculate the final posterior probability: The value of the numerator in the formula is... ;

[0088] The value of the denominator in the formula is ;

[0089] .

[0090] The results show that, given the observation sequence and model, the probability of the second observation occurring in the preparation phase is high. This is derived by comparing the posterior probability of the state (1.0) with the posterior probabilities of all other possible states (registration, injection, and confirmation phases) at the same time step (t=2), all of which are 0. This judgment is made because the probability value reaches the theoretical maximum, far exceeding all other candidate states. This calculation is performed for each observation in the sequence to obtain the posterior probability set for each phase state.

[0091] S213: For each stage state posterior probability group, the maximum a posteriori decision rule is used for each vaccination action to select the vaccination stage state with the largest posterior probability value and arrange them according to the order of vaccination operation time to generate a vaccination stage sequence.

[0092] For each observed behavior, a set of posterior probabilities for each stage state is calculated. For each set of posterior probabilities, the maximum a posteriori (MAP) decision rule is applied to determine the most likely vaccination stage state corresponding to that observed behavior. Specifically, for the first observed behavior in the sequence, the system compares its four corresponding posterior probability values: P(state = registration stage | observation sequence), P(state = preparation stage | observation sequence), P(state = injection stage | observation sequence), and P(state = confirmation stage | observation sequence). The vaccination stage state corresponding to the highest probability value is selected as the final determination state for that observed behavior. For example, for the second observed behavior (vaccine preparation, normal), its calculated posterior probability set might be: posterior probability of registration stage is 0.0011, posterior probability of preparation stage is 0.9988, posterior probability of injection stage is 0.0001, and posterior probability of confirmation stage is 0. Among these values, 0.9988 is the largest, corresponding to the state "preparation stage". Therefore, the system determines that the vaccination stage corresponding to this observed behavior is "preparation stage". The system independently performs this maximum a posteriori probability selection process once for each observation in the observation sequence. Each observation is thus assigned a unique, most probable vaccination stage status label. After all observations have been labeled, these determined vaccination stage states are arranged in ascending order strictly according to the chronological sequence of their corresponding original vaccination operations. This arrangement process ensures that the final generated sequence reflects the true timeline of the vaccination process. For example, in chronological order, the first observation is determined to be the "registration stage," the second to the "preparation stage," the third to the "injection stage," and the fourth to the "confirmation stage." Combining these determinations in sequence generates the final vaccination stage sequence.

[0093] Please see Figure 4 The specific steps for obtaining vaccination consistency records are as follows:

[0094] S311: Obtain the dose registration number and vaccine type code record of the recipient, call the vaccination stage sequence, match and associate the dose registration number and vaccine type code record with the status of each vaccination stage in the sequence, and establish a dose stage correspondence table;

[0095] First, two key identifying information for a specific vaccination service is retrieved from the vaccination management database: the recipient's dose registration number and the vaccine type code. For example, for a given vaccination, the retrieved dose registration number might be "D20251106HPV02" and the vaccine type code might be "HPVV9-01," indicating that this is the recipient's second nine-valent HPV vaccine. After obtaining this information, the vaccination stage sequence generated for this service in the preceding steps is invoked. This sequence is an ordered list, such as [Registration Stage, Preparation Stage, Injection Stage, Confirmation Stage]. Next, a matching and association operation is performed, binding the two macro-level identifying information—the dose registration number and the vaccine type code—to each micro-level vaccination stage status within the sequence. Essentially, this process creates a structured data record that not only contains the identification information for this vaccination but also a detailed breakdown of its internal processes. In practice, the system generates a data entity containing three fields: "Dose Number," "Vaccine Code," and "Stage Sequence." The obtained values ​​"D20251106HPV02," "HPVV9-01," and the stages [Registration Stage, Preparation Stage, Injection Stage, Confirmation Stage] are then filled into these fields. By generating such a record for each vaccination service and aggregating all records, a dose stage correspondence table is established, detailing the internal process status of each dose.

[0096] S312: For each dose in the dose-stage correspondence table, the sequence of vaccination stage states associated with each dose is compared with the flow of the vaccination stage state set item by item to verify the completeness of the vaccination stage state sequence and the accuracy of the order of arrangement, and to obtain the results of the stage order logic test.

[0097] For each record in the dose-stage correspondence table, a process involving dual checks of completeness and sequential accuracy is initiated for the associated vaccination stage status sequence. This check is based on a pre-defined, standard vaccination stage status set in a fixed order: [Registration Stage, Preparation Stage, Injection Stage, Confirmation Stage]. This standard procedure is also based on the national Standard Operating Procedure (SOP) for Vaccination, which specifies the unchangeable legal sequence of steps that vaccination services must follow. The first step of the check is a completeness check. The system compares the recorded vaccination stage status sequence, such as [Registration Stage, Injection Stage, Confirmation Stage], element-wise with the four stages in the standard procedure. By calculating the difference between the two sets, it determines whether a stage is missing. In the example above, the recorded sequence is missing the "Preparation Stage," so the completeness check result is recorded as "stage sequence incomplete." The second step of the check is a sequential accuracy check, which is only performed if the completeness check passes. If the recorded sequence contains all four stages, such as [Registration Stage, Preparation Stage, Confirmation Stage, Injection Stage], the system compares the elements of this sequence position by position with those of the standard procedure sequence. In the second position, the "Preparation Phase" matched successfully. However, in the third position, the recorded sequence was for the "Confirmation Phase," while the standard procedure was for the "Injection Phase," resulting in a mismatch. Once a mismatch is found at any position, the sequence accuracy check fails, and the result is recorded as "Incorrect Sequence." Only when the length of the recorded sequence is exactly the same as the standard sequence, and all corresponding elements are identical (e.g., [Registration Phase, Preparation Phase, Injection Phase, Confirmation Phase]), is the check result recorded as "Correct Sequence." By performing these two rigorous comparisons on each record, the system generates a clear logical check conclusion for each dose's vaccination procedure, obtaining the stage sequence logical check result.

[0098] S313: Based on the dose-stage correspondence table, calculate the time difference of vaccination operation between adjacent doses, compare the time difference of vaccination operation with the planned vaccination interval in the corresponding vaccine type code record, obtain the dose-stage time interval comparison result, integrate it with the stage sequence logic test result, and generate a vaccination consistency record.

[0099] Based on the established dose-stage correspondence table, the system calculates the vaccination operation time difference between the current dose and the previous dose. Specifically, the system first retrieves the recipient's last vaccination record for the same vaccine from the historical records using the recipient's unique identifier and vaccine type code in the current record. For example, if the current process involves the second dose of "HPVV9-01," the system will find the recipient's first dose vaccination record. Then, it extracts the vaccination operation timestamp for the first stage (registration stage) from the current second dose record and the vaccination operation timestamp for the last stage (confirmation stage) from the previous dose record. The difference between these two timestamps is calculated to obtain the vaccination operation time difference in days or months. Next, this calculated actual time difference is compared with the planned vaccination interval. The planned vaccination interval is not a fixed value but is derived by querying a built-in database derived from the officially published "National Immunization Program Vaccine Childhood Immunization Schedule" or the "Vaccine Instructions" provided by the vaccine manufacturer. The query is based on the vaccine type code record and the current dose. For example, for the second dose of "HPVV9-01", the database returns the recommended interval between it and the first dose, such as "2 months". During comparison, the system compares the calculated actual time difference, such as "65 days", with the planned vaccination interval of "2 months" (usually converted to 60 days). If the actual interval is greater than or equal to the planned interval, the comparison result is "compliant"; if it is less than the planned interval, the result is "incompatible". Finally, this dose interval comparison result is integrated with the stage sequence logic check result obtained in the previous step. The integration process merges the two results into a single record, forming a comprehensive vaccination consistency assessment. For example, a complete record will contain dose information, stage sequence, stage sequence logic check result (such as "correct sequence"), and dose interval comparison result (such as "compliant"), collectively generating a vaccination consistency record.

[0100] Please see Figure 5 The specific steps for obtaining the list of abnormal vaccination operations are as follows:

[0101] S411: Retrieve the comparison results reflected in the stage sequence logic test results within the vaccination consistency record, filter records with incomplete or incorrect stage sequence results, and generate an abnormal vaccination situation screening set if the time interval between doses within the record is less than the planned vaccination interval.

[0102] The entire set of vaccination consistency records is retrieved, and a screening operation is performed based on the comparison results reflected by the logical check of the stage sequence. The first screening criterion is to locate and extract all records with a comparison result of "incomplete stage sequence" or "incorrect sequence." This means that any vaccination record with missing steps or reversed step order will be selected. The second screening criterion is to find records among all records whose dose interval is shorter than the planned vaccination interval. This step is achieved by checking the "dose interval comparison result" field in each record and extracting records with a value of "non-compliant." Finally, records that meet either of the above criteria are grouped together. In other words, a record will be included in this set if its process logic is incorrect or its vaccination interval is too short. This subset obtained after screening with dual criteria is the abnormal vaccination situation screening set.

[0103] S412: For each record in the abnormal vaccination situation screening set, the abnormal content is classified into three preset abnormal vaccination types: missing stage order, disordered stage order, or abnormal dose interval. Type labels are added to the records to establish an abnormal vaccination type label library.

[0104] For each record in the abnormal vaccination screening set, the specific abnormal content is analyzed and categorized into one of three preset abnormal vaccination types. These three preset types are: missing stage sequence, out-of-order stage sequence, and abnormal dose interval. Step S312 checks the completeness and sequence of the process, with failure modes corresponding to "missing stage sequence" and "out-of-order stage sequence"; step S313 checks the temporal relationship between processes, with failure mode corresponding to "abnormal dose interval". The categorization process follows explicit rules: if a record's "stage sequence logical check result" is "incomplete stage sequence", it is first categorized as "missing stage sequence" regardless of whether its time interval is abnormal. If a record's "stage sequence logical check result" is "incorrect sequence", it is categorized as "out-of-order stage sequence". If a record's "stage sequence logical check result" is "correct sequence", but its "dose interval comparison result" is "incompatible", it is categorized as "abnormal dose interval". After determining the abnormal vaccination type for each abnormal record, the system adds a type label to that record. This label is a new data field whose value is one of "missing stage order", "out of order of stages", or "abnormal dose interval". By adding such a type label to all records in the filter set, the original filter set is transformed into a more clearly structured and information-rich abnormal vaccination type label library.

[0105] S413: Based on the abnormal vaccination type labeling library, extract the unique identifier of the recipient, the device identification number and the operator identification number corresponding to each record, and group and arrange them according to the labeled abnormal vaccination types to construct an abnormal vaccination operation list;

[0106] The system iterates through each marked abnormal record in the database, extracting three key identifying information from each record: the recipient's unique identifier, the device identification number, and the operator's identifier. This information indicates the responsible party and the context of the abnormal operation. After extracting the necessary information, the records are grouped and arranged according to the marked abnormal vaccination type. Specifically, the system creates three lists corresponding to the three types: "Missing Phase Sequence," "Out of Order Phase Sequence," and "Abnormal Dosage Interval." Then, each record with extracted identifiers is placed into the corresponding list based on its type. For example, the identifier information of all records marked "Missing Phase Sequence" will be stored in the "Missing Phase Sequence" group. Through this grouping and arrangement, a structured and clearly categorized list of abnormal vaccination operations is finally constructed.

[0107] Please see Figure 6 The specific steps for obtaining the verification results of the vaccination process are as follows:

[0108] S511: For each abnormal vaccination type in the abnormal vaccination operation list, extract the corresponding posterior probability from the posterior probability group of each stage state, and perform aggregate operation on the posterior probabilities of all abnormal operations under each abnormal vaccination type to obtain the probability confidence of abnormal vaccination operation.

[0109] Read the type label field in the abnormal vaccination operation list, and group all records into corresponding record groups according to "missing stage order", "out of order stage", and "abnormal dose interval". Then, read the abnormal occurrence location, stage time index, and abnormal association number for each abnormal record, and look up the probability value of the record at the same time position in the stage status posterior probability group according to the abnormal association number. When the type label is "missing stage order", extract the posterior probability corresponding to the actual stage before the missing position and the actual stage after the missing position, and extract the posterior probability of the missing stage at that time position. Calculate the difference between the posterior probability of the missing stage and the posterior probabilities of the adjacent actual stages. The absolute value is used to calculate the missing bias value. Then, the posterior probability of the missing stage is compared with the mean of the posterior probabilities of the adjacent actual stages. If the posterior probability of the missing stage is less than the mean, the missing stage posterior probability is retained as the anomaly probability value of the anomaly record. If the posterior probability of the missing stage is equal to the mean, both the missing stage posterior probability and the missing bias value are recorded. If the posterior probability of the missing stage is greater than the mean, the deviation of the missing stage posterior probability from the mean is recorded. When the type is labeled as stage order disorder, the posterior probabilities of the previous stage, the current stage, and the next stage under the same time index are extracted according to the location of the anomaly. First, the posterior probability of the current stage is calculated and compared with the standard order. The ratio of the posterior probability of the correct position stage is used to calculate the difference between the posterior probability of the current stage and the posterior probabilities of the previous and next stages. If the posterior probability of the current stage is greater than the posterior probability of the correct position stage, the difference between the two probabilities is recorded as a disordered offset value. If the posterior probability of the current stage is equal to the posterior probability of the correct position stage, both the current stage posterior probability and the posterior probability of the correct position stage are recorded as parallel values. If the posterior probability of the current stage is less than the posterior probability of the correct position stage, the posterior probability of the correct position stage is used as the verification probability value for this abnormal record. When the type is labeled as abnormal dose interval, the first stage state of the dose associated with the abnormal dose record is extracted and then compared with... The posterior probability under the time index is calculated, and the posterior probability of the last stage state of the previous dose under the corresponding time index is extracted. The mean of the posterior probability of the first stage of the current dose and the posterior probability of the last stage of the previous dose is calculated as the base probability value of the interval abnormal record. After the extraction of a single record is completed, within each abnormal vaccination type, the abnormal probability values ​​of all abnormal records are aggregated. First, the ratio of the sum of all abnormal probability values ​​of the same type to the total number of abnormal records is calculated to obtain the type mean. Then, the ratio of the sum of the absolute values ​​of the deviations between each abnormal probability value and the type mean to the total number of abnormal records is calculated to obtain the discrete metric. Finally, the difference between the type mean and the discrete metric is calculated to obtain the initial confidence value of the type.If the initial confidence value is less than zero, the confidence level of the abnormal vaccination operation probability is recorded as zero; if the initial confidence value is greater than one, the confidence level of the abnormal vaccination operation probability is recorded as one; if the initial confidence value is between zero and one, it is directly recorded as the confidence level of the abnormal vaccination operation probability. When there is only one abnormal record under the same abnormality type, the abnormality probability value of that abnormal record is directly read as the confidence level of the abnormal vaccination operation probability. When there are two or more abnormal records under the same abnormality type, the calculation is completed in the order of the type mean, dispersion measure, and initial confidence value, and the corresponding values ​​are output to form the confidence level of the abnormal vaccination operation probability corresponding to that abnormal vaccination type.

[0110] S512: Set up multi-level confidence intervals, compare the probability confidence of each type of abnormal vaccination operation with the confidence interval, and assign one of the three types of early warning instructions (time out of bounds, stage disorder, and dose skipping) according to the range of the interval, and obtain a graded early warning instruction set;

[0111] A multi-level confidence interval is established, consisting of multiple non-overlapping intervals. These intervals are determined through statistical analysis of a dataset of confidence scores generated from a large number of historically verified abnormal vaccination procedures, after calculating their probability confidence levels. Specifically, by analyzing the probability density distribution of this dataset, the cutoff points that effectively distinguish abnormal events of different risk levels are identified. For example, the analysis shows that anomalies with confidence levels above a certain high quantile (e.g., the 90th percentile) are mostly data recording errors and have low risk; anomalies with confidence levels between the median and high quantiles often involve process irregularities and have moderate risk; and anomalies with confidence levels below the median quantile typically involve serious process skipping or time violations and have high risk. Therefore, three confidence intervals can be set: (0.9, 1.0], (0.7, 0.9], and [0, 0.7]. The confidence level of the probability of each type of abnormal vaccination operation is compared with these confidence intervals. Based on the range in which the value falls, a preset warning instruction is assigned. These three types of warning instructions are time out of bounds, stage disorder, and dose skipping, which are logical mappings based on risk level and abnormality type. The mapping rule is as follows: if the confidence level falls into the high interval (0.9, 1.0], it indicates a weak abnormal signal, and no warning instruction is generated; if it falls into the middle interval (0.7, 0.9], it indicates a logical error in the process, and a "stage disorder" warning instruction is assigned; if it falls into the low interval [0, 0.7], it indicates a serious violation, and in this case, a more targeted "time out of bounds" or "dose skipping" warning instruction is assigned based on the specific type of abnormality (time-related or step-related). Through this comparison and assignment process, a specific warning level is calculated for each type of risk, and a graded warning instruction set is obtained.

[0112] S513: Call the graded early warning instruction set, extract the unique identifier of the vaccine recipient from the abnormal vaccination operation list, obtain the dose registration number from the dose stage correspondence table, integrate the early warning instruction with the unique identifier of the vaccine recipient and the dose registration number, and generate the vaccine vaccination process verification result.

[0113] The system invokes a tiered early warning instruction set and integrates information from the abnormal vaccination operation list and the dose stage correspondence table to perform final verification results. Specifically, the system iterates through each valid instruction in the tiered early warning instruction set. For an instruction, such as an "out-of-order stage" warning, the system searches the abnormal vaccination operation list and filters out all records of this type. For each filtered abnormal record, the system extracts the recipient's unique identifier. Simultaneously, using this identifier and vaccination time information, the system queries the dose stage correspondence table to obtain the corresponding dose registration number. Finally, the early warning instruction, the recipient's unique identifier, and the dose registration number are integrated to generate a structured vaccination process verification result record. For example, a result record might look like this: "Early Warning Instruction: Out-of-order Stage; Recipient Identifier: PID20251106001; Dose Number: D20251106HPV02". This integration step is performed on all abnormal operations requiring an early warning, ultimately generating the vaccination process verification result.

[0114] An intelligent verification system for a vaccination process, the intelligent verification system for a vaccination process being used to execute the aforementioned intelligent verification method for a vaccination process, the system comprising:

[0115] The vaccination behavior data acquisition module acquires information about vaccine recipients and vaccination implementers during the vaccination process, matches the information with a preset vaccination management database, determines homologous vaccination behaviors, and constructs a vaccine recipient vaccination behavior sequence.

[0116] The vaccination stage inference module inputs the recipient's vaccination behavior sequence into the Hidden Markov Model, calculates the posterior probability of each vaccination behavior under the preset vaccination stage state set, arranges each vaccination stage state with the highest posterior probability in time, and generates the vaccination stage sequence.

[0117] The dose-stage consistency verification module obtains the recipient's dose registration number and vaccine type code record, establishes a dose-stage correspondence table according to the vaccination time sequence, verifies the order of each vaccination stage status in the vaccination stage sequence based on the dose-stage correspondence table, and obtains the vaccination consistency record.

[0118] The abnormal vaccination identification module filters abnormal vaccination cases based on vaccination consistency records, marks the abnormal vaccination type, and constructs a list of abnormal vaccination operations.

[0119] The process verification and archiving module determines the probability confidence level of each type of abnormal vaccination operation in the abnormal vaccination operation list, assigns early warning instructions based on the confidence level interval, and obtains the vaccination process verification results.

[0120] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for intelligent verification of vaccination process characterized in that, Includes the following steps: S1: Obtain information about the vaccine recipient and the vaccination implementer during the vaccination process, match the information with the preset vaccine management database, determine the same-origin vaccination behavior, and construct the vaccine recipient's vaccination behavior sequence; S2: Input the recipient's vaccination behavior sequence into a hidden Markov model, calculate the posterior probability of each vaccination behavior under a preset vaccination stage state set, arrange the vaccination stage states with the highest posterior probability in time, and generate a vaccination stage sequence. S3: Obtain the recipient's dose registration number and vaccine type code record, establish a dose stage correspondence table according to the vaccination time sequence, verify the order of each vaccination stage status in the vaccination stage sequence according to the dose stage correspondence table, and obtain the vaccination consistency record; S4: Based on the vaccination consistency record, filter abnormal vaccination cases, mark the abnormal vaccination type, and construct an abnormal vaccination operation list; S5: Determine the probability confidence level of each type of abnormal vaccination operation in the abnormal vaccination operation list, allocate early warning instructions according to the confidence level interval, and obtain the vaccination process verification results.

2. The method for intelligent verification of vaccination process as claimed in claim 1 wherein, The recipient vaccination behavior sequence includes a vaccination behavior identifier, a behavior time index, and a behavior type label. The vaccination stage sequence includes a stage status name, a stage time index, and a corresponding probability value for each stage. The vaccination consistency record includes a stage sequence structure, a dose correspondence, and a time connection status. The abnormal vaccination operation list includes an abnormality type, an abnormality location, and an abnormality association number. The vaccination process verification result includes an abnormality probability confidence interval, a warning instruction level, and a recipient corresponding identifier.

3. The method for intelligent verification of vaccination process as claimed in claim 1 wherein, The specific steps for obtaining the recipient's vaccination behavior sequence are as follows: S111: Obtain information about the vaccine recipient and the vaccine executor during the vaccination process, including the recipient's unique identifier, device identification number, operator identification number, vaccine operation type and vaccine operation time. Based on the recipient's unique identifier, filter all corresponding vaccination log records in the preset vaccine management database, determine the consistency between the device identification number and the operator identification number, and establish the same source vaccination behavior determination result. S112: Based on the determination result of the same vaccination behavior, all vaccination log records of the same recipient are aggregated, and the vaccination operation type corresponding to each operation event is matched with the preset vaccination event type, including information registration, vaccine preparation, vaccine injection and result confirmation, to obtain the aggregated vaccination event type mapping set. S113: Based on the aggregated vaccination event type mapping set, all vaccination operations for the same recipient are arranged in ascending order of vaccination operation time, and each operation event is treated as a vaccination behavior record to generate a recipient vaccination behavior sequence.

4. The method for intelligent verification of vaccination process as claimed in claim 3, wherein, The specific steps for obtaining the vaccination phase sequence are as follows: S211: Define the vaccination operation type of each vaccination behavior record in the vaccination behavior sequence of the recipient as an observation variable, calculate the time difference between adjacent vaccination operation times as a time feature, and input the observation variable and time feature into a hidden Markov model to establish a vaccination behavior observation feature set. S212: Based on the observation feature set of the vaccination behavior and the initial state distribution and state transition probability of the hidden Markov model, for the preset vaccination stage state set, including the registration stage, preparation stage, injection stage and confirmation stage, calculate the posterior probability of each observation behavior in each vaccination stage state, and obtain the posterior probability set of each stage state. S213: For each stage state posterior probability group, the maximum a posteriori decision rule is applied to each vaccination action to select the vaccination stage state with the largest posterior probability value and arrange them according to the vaccination operation time sequence to generate a vaccination stage sequence.

5. The method for intelligent verification of vaccination process as claimed in claim 4, wherein, The process of combining observed variables and time features and inputting them into a Hidden Markov Model to establish a set of observed features of vaccination behavior is as follows: Multiple time thresholds are preset, and the time thresholds are determined based on the statistical distribution of historical vaccination operation time data; The time features are quantized into intervals based on time thresholds and mapped to discrete time interval labels. The vaccination operation type of each vaccination behavior record in the vaccination behavior sequence is taken as a discrete observation category, and a binary observation group is formed by the time interval label mapped by the time feature of each vaccination behavior record. The observation probability of the Hidden Markov Model is calculated based on the binary observation group. All binary observation groups are arranged in chronological order of vaccination operation to form an observation sequence, which is the observation feature set of vaccination behavior.

6. The method for intelligent verification of vaccination process as claimed in claim 4 wherein, The specific steps for obtaining the vaccination consistency record are as follows: S311: Obtain the dose registration number and vaccine type code record of the recipient, call the vaccination stage sequence, match and associate the dose registration number and vaccine type code record with the status of each vaccination stage in the sequence, and establish a dose stage correspondence table; S312: For each dose in the dose stage correspondence table, the vaccination stage state sequence associated with each dose is compared with the flow of the vaccination stage state set item by item to check the completeness of the vaccination stage state sequence and the accuracy of the order of arrangement, and to obtain the stage order logic check result. S313: Based on the dose-stage correspondence table, calculate the time difference of vaccination operation between adjacent doses, compare the time difference of vaccination operation with the planned vaccination interval in the corresponding vaccine type code record to obtain the dose-stage time interval comparison result, integrate it with the stage sequence logic test result, and generate a vaccination consistency record.

7. The method for intelligent verification of vaccination procedures as claimed in claim 6, wherein, The specific steps for obtaining the abnormal vaccination operation list are as follows: S411: Retrieve the comparison results reflected in the stage sequence logic test results within the vaccination consistency record, filter records with incomplete or incorrect stage sequence results, and generate an abnormal vaccination situation screening set by recording the time interval between doses that is less than the planned vaccination interval. S412: For each record in the abnormal vaccination situation screening set, the abnormal content is classified into three preset abnormal vaccination types: missing stage order, disordered stage order, or abnormal dose interval. Type labels are added to the records to establish an abnormal vaccination type label library. S413: Based on the abnormal vaccination type labeling library, extract the unique identifier of the recipient, the device identification number, and the operator's identifier corresponding to each record, and group and arrange them according to the labeled abnormal vaccination types to construct an abnormal vaccination operation list.

8. The method for intelligent verification of vaccination procedures as claimed in claim 7, wherein, The specific steps for obtaining the verification results of the vaccination process are as follows: S511: For each abnormal vaccination type in the abnormal vaccination operation list, extract the corresponding posterior probability from the posterior probability group of each stage state, and perform aggregate operation on the posterior probabilities of all abnormal operations under each abnormal vaccination type to obtain the probability confidence of abnormal vaccination operation. S512: Set up multi-level confidence intervals, compare the probability confidence of each type of abnormal vaccination operation with the confidence interval, and assign one of the three types of early warning instructions (time out of bounds, stage disorder, and dose skipping) according to the range of the interval, and obtain a graded early warning instruction set; S513: Invoke the tiered early warning instruction set, extract the recipient's unique identifier from the abnormal vaccination operation list, obtain the dose registration number from the dose stage correspondence table, integrate the early warning instruction with the recipient's unique identifier and dose registration number, and generate the vaccination process verification result.

9. An intelligent verification system for a vaccination process, characterized in that, The intelligent verification method for a vaccination process according to any one of claims 1-8, wherein the system comprises: The vaccination behavior data acquisition module acquires information about vaccine recipients and vaccination implementers during the vaccination process, matches the information with a preset vaccination management database, determines homologous vaccination behaviors, and constructs a vaccine recipient vaccination behavior sequence. The vaccination stage inference module inputs the recipient's vaccination behavior sequence into a hidden Markov model, calculates the posterior probability of each vaccination behavior under a preset vaccination stage state set, arranges each vaccination stage state with the highest posterior probability in time, and generates a vaccination stage sequence. The dose-stage consistency verification module obtains the recipient's dose registration number and vaccine type code record, establishes a dose-stage correspondence table according to the vaccination time sequence, verifies the order of each vaccination stage status in the vaccination stage sequence based on the dose-stage correspondence table, and obtains the vaccination consistency record. The abnormal vaccination identification module filters abnormal vaccination cases based on the vaccination consistency record, marks the abnormal vaccination type, and constructs an abnormal vaccination operation list. The process verification and archiving module determines the probability confidence level of each type of abnormal vaccination operation in the abnormal vaccination operation list, allocates early warning instructions according to the confidence level interval, and obtains the vaccination process verification results.

Images