Open queue construction method and system with dynamic balance mechanism

Abstract

The invention relates to an open type queue construction method and system with a dynamic balance mechanism, belongs to the technical field of medical health research, and solves the problem of bias of an existing open type queue method. The method comprises the following steps: determining out-of-visit individuals in the current period in each time period of each period of a follow-up visit process; filling the data of the out-of-visit individuals based on the follow-up visit data of the in-queue individuals; based on the filling data of the out-of-visit individuals, matching substitute individuals for the out-of-visit individuals in the to-be-incorporated sub-population by adopting coarse granularity accurate matching, and adding the successfully matched substitute individuals into a queue; and if the unmatched visit-losing individuals exist, determining hierarchical variable distribution of the unmatched visit-losing individuals based on the filling data of the unmatched visit-losing individuals, searching substitute individuals of the unmatched visit-losing individuals in the to-be-included crowd based on the hierarchical variable distribution of the unmatched visit-losing individuals, and adding the substitute individuals into a queue. The balance of variable differences between the out-of-visit crowd and the newly-added crowd is realized, and the external validity of queue research is improved.

Description

Technical Field

[0001] The present invention relates to the field of medical health research technology, and in particular to an open queue construction method and system with a dynamic balance mechanism. Background Art

[0002] Although large population closed cohort studies can capture the impact of changes in life and environment on population health, they are often difficult to adapt to the needs of high population mobility in modern research, resulting in a large number of lost people in the follow-up process, making it difficult to fully reflect the overall health trend of the population. Open cohort studies provide a new way to solve this problem, but the existing open cohort methods have limitations in the inclusion of new participants, especially in the lack of effective strategies for dealing with the differences in variables between new participants and lost participants, making them unsuitable for large population cohort studies that have high requirements for the representativeness of the research population.

[0003] In an open cohort study, the target population is established at baseline and new participants are included during follow-up until the end of the study. In traditional open cohort studies, the follow-up population lacks representativeness of the target population due to the significant differences between new participants and lost participants during follow-up. This significant difference involves both time-invariant and time-varying variables, thereby reducing the external validity of the cohort study. Take the gender and average age of the population, a key baseline covariate in a large population cohort survey, as an example. The gender of the cohort population does not change over time, while the average age increases with follow-up. In the case of new participants, it may be the case that the lost population has a lower proportion of women and a higher age, while the new population has a higher proportion of women and a lower age. This imbalance of variables between the lost population and the new population will affect the investigation of trends in health risk factors and the evaluation of prevention strategies, thereby affecting the reliability and credibility of the conclusions.

[0004] In traditional open cohort studies, researchers usually use a closed cohort method during the data analysis phase, excluding new participants in follow-up or using the length of time in the cohort to adjust the proportion of individual contributions in data analysis. However, this faces the risk of sample bias caused by inconsistency with the target population. Summary of the invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide an open queue construction method and system with a dynamic balance mechanism, so as to solve the problem of bias in the existing open queue method.

[0006] On the one hand, an embodiment of the present invention provides an open queue construction method with a dynamic balancing mechanism, comprising the following steps:

[0007] In each period of each phase of the follow-up process, the individuals lost to follow-up in the current period were identified;

[0008] The data of lost individuals were imputed based on the follow-up data of individuals in the team;

[0009] Based on the imputed data of lost individuals, coarse-grained precise matching is used to match the lost individuals with replacement individuals in the sub-population to be included, and the successfully matched replacement individuals are added to the cohort;

[0010] If there are unmatched lost individuals, the stratification variable distribution of the unmatched lost individuals is determined based on the imputed data of the unmatched lost individuals, and replacement individuals for the unmatched lost individuals are found in the population to be included based on the stratification variable distribution of the unmatched lost individuals to join the cohort.

[0011] Based on the further improvement of the above method, the data of lost individuals are filled in based on the follow-up data of the individuals in the team, including:

[0012] The first type of variables for individuals lost to follow-up were directly imputed;

[0013] For the second type variables of the lost individuals, the propensity score matching method was used to match similar individuals to the lost individuals from the individuals in the team, and the second type variables of the lost individuals were filled based on the second type variable values ​​of the similar individuals;

[0014] Among them, the first type of variables are variables that do not change with time or that change with time but the change pattern is known, and the second type of variables are variables that change with time and the change pattern is unknown.

[0015] Based on the further improvement of the above method, the following formula is used to fill the second type of variables of the lost individuals based on the variable values ​​corresponding to similar individuals:

[0016]

[0017] Xk represents the kth second type variable, represents the value of the kth second type variable of the mth similar individual in the tth period, M0de(·) represents the calculated mode, M represents the number of similar individuals, It represents the imputed value of the kth second-type variable of the i-th individual at the t-th period, and the i-th individual is the lost individual.

[0018] Based on the further improvement of the above method, the propensity score of each individual lost to follow-up in period t is calculated based on the following formula, and the M individuals with the closest propensity score to the lost individual are selected from the individuals in the team as similar individuals:

[0019]

[0020] in, represents the propensity score of the i-th individual being lost to follow-up in period t, indicates that the i-th individual was lost to follow-up in period t, represents the second type variable of the i-th individual at the t-1 period, represents the first type variable of the ith individual at the t-1th period, and P(·) represents the probability.

[0021] Based on the further improvement of the above method, coarse-grained precise matching is used to match replacement individuals for lost individuals in the sub-population to be included, including:

[0022] Stratification was performed for each non-categorical variable at follow-up, and stratification remained unchanged for each categorical variable at follow-up;

[0023] For each individual lost to follow-up, the stratum of each variable of the individual lost to follow-up was determined based on the imputed data of each variable of the individual lost to follow-up;

[0024] If there is an individual in the sub-population to be included whose stratification variable distribution is consistent with that of the lost individual, the matching is successful, and an individual with a consistent stratification variable distribution is randomly selected as a replacement individual for the lost individual.

[0025] Based on the further improvement of the above method, the method further includes:

[0026] If the current time period is not the last time period of the current period, the target new population is searched in the population to be included based on the filling data of the lost individuals accumulated in the current period, and the target new population is added to the sub-population to be included.

[0027] On the other hand, an embodiment of the present invention provides an open queue construction system with a dynamic balancing mechanism, comprising the following modules:

[0028] A lost-follow-up individual determination module is used to determine the lost-follow-up individual in the current period in each period of each period of the follow-up process;

[0029] A data filling module is used to fill in the data of lost individuals based on the follow-up data of individuals in the team;

[0030] The precise matching module is used to match the missing individuals with replacement individuals in the sub-population to be included based on the filling data of the missing individuals, using coarse-grained precise matching, and adding the successfully matched replacement individuals to the cohort;

[0031] The distribution matching module is used to determine the stratified variable distribution of the unmatched lost individuals based on the filled data of the unmatched lost individuals if there are any unmatched lost individuals, and to find replacement individuals for the unmatched lost individuals in the population to be included based on the stratified variable distribution of the unmatched lost individuals to add to the queue.

[0032] Based on the further improvement of the above system, the data of lost individuals are filled in based on the follow-up data of the individuals in the team, including:

[0033] The first type of variables for individuals lost to follow-up were directly imputed;

[0034] For the second type variables of the lost individuals, the propensity score matching method was used to match similar individuals to the lost individuals from the individuals in the team, and the second type variables of the lost individuals were filled based on the second type variable values ​​of the similar individuals;

[0035] Among them, the first type of variables are variables that do not change with time or that change with time but the change pattern is known, and the second type of variables are variables that change with time and the change pattern is unknown.

[0036] Based on the further improvement of the above system, the following formula is used to fill the second type of variables of the lost individuals based on the variable values ​​corresponding to similar individuals:

[0037]

[0038] X ·k represents the kth second type variable, represents the value of the kth second type variable of the mth similar individual at the tth period, Mode(·) represents the calculation of the mode, M represents the number of similar individuals, It represents the imputed value of the kth second-type variable of the i-th individual at the t-th period, and the i-th individual is the lost individual.

[0039] Based on the further improvement of the above system, the system further includes:

[0040] The strategy adjustment module is used to find the target new population in the population to be included based on the filling data of the lost individuals accumulated in the current period if the current period is not the last period of the current period, and add the target new population to the sub-population to be included.

[0041] Compared with the prior art, the present invention fills in the data of lost individuals based on the follow-up data of individuals in the cohort, thereby reducing the bias with the target population set at the initial stage of the study and enhancing the representativeness of the target population, matching lost individuals with substitute individuals with consistent distribution of stratification variables through coarse-grained precise matching, searching for substitute individuals of unmatched lost individuals in the population to be included based on the distribution of stratification variables of unmatched lost individuals in the non-final time period, that is, searching for individuals with consistent distribution of stratification variables with unmatched lost individuals as substitute individuals, thereby reducing the bias between substitute individuals and lost individuals, dynamically balancing the follow-up variables of the lost population and the newly added population, improving the external validity of the cohort study, and improving the reliability and credibility of the research results.

[0042] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be achieved and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0044] Figure 1 A flowchart of a method for constructing an open queue with a dynamic balancing mechanism according to an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of an open queue according to an embodiment of the present invention;

[0046] Figure 3 This is a comparison diagram of the effects of a single simulation of an embodiment of the present invention;

[0047] Figure 4 This is a comparison diagram of the effects of multiple simulations of an embodiment of the present invention;

[0048] Figure 5 A block diagram of an open queue construction system with a dynamic balancing mechanism according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0050] Consider a cohort study with a duration of T+1, where the cohort individuals come from the set A series of variables are collected for each individual, including K variables that change over time and whose changing patterns are unknown (such as height, body mass index, etc.) and L variables that do not change over time or change over time but whose changing patterns are known (such as gender, age, etc.). The team status of the i-th individual in the tth period is recorded as In the team, Represents not in the team. Use N t represents the number of people in the queue in period t, t=0,…,T, then The observed variable of the ith individual in period t is t=0,…,T,i=1,…,N t .in, Variables that do not change over time or that change over time but the pattern of change is known are called first-class variables; Variables that change over time and whose changing patterns are unknown are called second-class variables. Consider the following open queue:

[0051] At the beginning of the cohort recruitment, all individuals were not in the cohort and were

[0052] When t = 0, N 0 Individuals enter the queue, record the individuals entering the queue

[0053] When t=1,…,T, for the group of people in the previous period, that is, The follow-up of a large population is usually divided into multiple periods (denoted as J), and each period follows up a part of the individuals in the population, that is, in each period, the current period follow-up is completed through multiple periods.

[0054] During the follow-up process, there may be individuals from the previous period who are lost to follow-up in the current period, so new individuals need to be included.

[0055] In the last period of period t, i.e. period J, the number of individuals currently in the team is N. t The open queue with T = 2 periods and J = 3 periods is as follows Figure 2 Hint.

[0056] In order to effectively balance these variables between the lost population and the newly added population, a specific embodiment of the present invention discloses an open queue construction method with a dynamic balance mechanism, such as Figure 1 As shown, the following steps are included:

[0057] S1. In each period of each period of the follow-up process, determine the individuals lost to follow-up in the current period;

[0058] S2, fill in the data of lost individuals based on the follow-up data of individuals in the team;

[0059] S3. Based on the imputation data of the lost individuals, coarse-grained precise matching is used to match the lost individuals with replacement individuals in the sub-population to be included, and the successfully matched replacement individuals are added to the cohort;

[0060] S4. If there are unmatched lost individuals, the stratified variable distribution of the unmatched lost individuals is determined based on the filled-in data of the unmatched lost individuals, and replacement individuals of the unmatched lost individuals are found in the population to be included based on the stratified variable distribution of the unmatched lost individuals to add to the cohort.

[0061] It should be noted that the individuals lost in the current period are individuals who were in the team in the previous period but are not in the team in the current period.

[0062] In the cohort follow-up, the population to be included is a group of individuals who meet the recruitment conditions but have not yet entered the cohort. Initially, the individuals to be included in the population to be included who are in the same location (geographic location, education, culture and economic development level) as the cohort implementation site constitute the sub-population to be included.

[0063] Compared with the prior art, the open cohort construction method with a dynamic balance mechanism provided in this embodiment fills in the data of lost individuals based on the follow-up data of individuals in the cohort, thereby reducing the bias with the target population set at the initial stage of the study and enhancing the representativeness of the target population. The lost individuals are matched with substitute individuals with consistent distribution of stratification variables through coarse-grained precise matching. In the non-final period, substitute individuals for unmatched lost individuals are searched in the population to be included based on the distribution of stratification variables of unmatched lost individuals, that is, individuals with consistent distribution of stratification variables of unmatched lost individuals are searched as substitute individuals, thereby reducing the bias between substitute individuals and lost individuals, dynamically balancing the follow-up variables of the lost population and the newly added population, improving the external validity of the cohort study, and improving the reliability and credibility of the research results.

[0064] Specifically, the data of lost individuals are filled based on the follow-up data of the individuals in the team, including:

[0065] S21, directly fill in the first type of variables for the individuals who were lost to follow-up;

[0066] S22. For the second type variable of the lost individual, a propensity score matching method is used to match similar individuals for the lost individual from the individuals in the team, and the second type variable of the lost individual is filled based on the second type variable value of the similar individual;

[0067] Among them, the first type of variables are variables that do not change with time or that change with time but the change pattern is known, and the second type of variables are variables that change with time and the change pattern is unknown.

[0068] Follow-up variables may include the first type of variables that do not change over time or change over time but the change pattern is known, such as gender, age, etc., and may also include variables that change over time and the change pattern is unknown, such as body mass index, blood pressure, etc. For the first type of variables, they can be filled directly according to the baseline and the change pattern. The baseline is the variable value at the initial state of the individual.

[0069] We assume that, given the variables of the previous period of the lost individuals, whether the lost individuals will be lost is irrelevant to the current period value of the observed variables. That is: t=1,…,T. For each i-th individual lost to follow-up in period t, the propensity score matching method is used to match M (for example, M=50) individuals in the cohort (repeatable) for the lost individual as similar individuals, and the second type variables of the i-th individual lost to follow-up are filled based on the second type variables of the matched M individuals.

[0070] The propensity score of the i-th individual being lost to follow-up in period t is in, indicates that the i-th individual was lost to follow-up in period t, represents the second type variable of the i-th individual at the t-1 period, represents the first type variable of the i-th individual in period t-1, and P(·) represents the probability. The propensity score means the probability that an individual will be lost to follow-up in the current period, given the known follow-up variables of the previous period. For each individual lost to follow-up in period t, the propensity score matching method is used to match the lost individual with M (for example, M=50) in-cohort individuals (repeatable) with the propensity score closest to the lost individual as similar individuals.

[0071] Each second-type variable of the lost individuals was then filled in sequentially, depending on whether the variable was a categorical variable.

[0072] For example, sex is a categorical variable, but body mass index is not.

[0073] Specifically, the following formula is used to fill in the second type of variables of the lost individuals based on the variable values ​​corresponding to similar individuals:

[0074]

[0075] X ·k represents the kth second type variable, represents the value of the kth second type variable of the mth similar individual at the tth period, Mode(·) represents the calculation of the mode, M represents the number of similar individuals, It represents the imputed value of the kth second-type variable of the i-th individual at the t-th period, and the i-th individual is the lost individual.

[0076] For the second type of categorical variables, the mode of similar individuals is calculated and the mode is taken as the variable value of the lost individuals. For the second type of non-categorical variables, the average value of similar individuals is taken as the variable value of the lost individuals.

[0077] Due to missing data, it is impossible to know the data characteristics of the lost population. By filling in the data of the lost individuals in the current period, we can prepare for the next step of finding replacement individuals for the lost individuals from the sub-population to be included based on the data characteristics.

[0078] After filling in the variables of the lost population, the coarse-grained and precise (1:1) matching method was used to match the variables of the sub-population to be included and the lost population. The successfully matched individuals to be included became the substitutes for the lost individuals.

[0079] Specifically, coarse-grained precise matching is used to match replacement individuals for lost individuals in the sub-population to be included, including:

[0080] Stratification was performed for each non-categorical variable at follow-up; stratification remained unchanged for each categorical variable at follow-up;

[0081] For each individual lost to follow-up, the stratum of each variable of the lost individual is determined based on the imputed data of each variable of the lost individual;

[0082] If there is an individual in the sub-population to be included with the same distribution of the stratification variable as the lost individual, the match is successful, and an individual with the same distribution of the stratification variable is randomly selected as a replacement individual for the lost individual. During implementation, the categorical variables in the follow-up variables are kept unchanged, that is, each type of categorical variable naturally forms a stratum. For each non-categorical variable, determine the split point and stratify it. For example, if there are n split points, n+1 intervals are formed, that is, n+1 layers.

[0083] During implementation, the split points can be evenly determined based on the value range of the non-categorical variable and the number of split points n. For example, in a cohort with an age range of 10-90 years old, a split point for the age variable is determined every 10 years or every 20 years, and the interval size of the split points can be determined based on the number of people in the cohort.

[0084] For each lost individual, the stratum of each variable of the lost individual is determined based on the imputed data of each variable in the current period, and one individual (if any) with consistent distribution of stratified variables in the sub-population to be included is randomly selected as the final replacement subject to form a 1:1 match. For example, there are 20 follow-up variables in total. After determining the stratum of the imputed data of each variable of the lost individual on the variable, find out from the sub-population to be included whether there is an individual with consistent variable distribution, that is, an individual with the same stratum of each variable value on the variable as the lost individual. If so, one is randomly selected as the replacement individual for the lost individual.

[0085] The sub-population to be included may not contain individuals with the same distribution of stratification variables as the lost individuals, and the precise matching of the lost individuals will fail.

[0086] If there are unmatched individuals who have been lost to follow-up, replacement individuals will be sought from the population to be included for the unmatched individuals.

[0087] Specifically, firstly, the distribution of stratified variables of unmatched lost individuals is determined based on the imputed data of unmatched lost individuals, and then, according to the distribution of stratified variables of unmatched lost individuals, individuals with corresponding distribution are found from the population to be included as substitute individuals to join the cohort. For example, the distribution table of stratified variables of unmatched lost individuals can be shown in Table 1, where the number of second type variables K=2, the number of first type variables L=2, and the first second type variable There are 6 layers in total, the second second type variable There are 9 layers in total, the first type of variable There are 2 layers in total, the second one is the first type variable There are 3 strata in total. Among the unmatched individuals who were lost to follow-up, 78 individuals had stratification variables distributed as the second type of variables. The first stratum second type variable The eighth stratum, the first type 1 variable The second stratum, the second first type variable 78 individuals with the same or most similar distribution of stratification variables were selected from the population to be included as replacement individuals for the 78 lost individuals and added to the cohort.

[0088] Table 1. Distribution of stratified variables in the unmatched population

[0089]

[0090] After matching the corresponding replacement individuals for the lost individuals, if the current period is not the last period, the inclusion strategy needs to be adjusted to increase the probability of successful matching of the lost individuals in the next period. Assuming that the lost populations in different periods of the same period are similar, the variable distribution of the cumulative lost population is most likely to be similar to the lost population in the next period, so an external population with the same variable distribution as the cumulative lost population is found as the new population, so that the variable distribution of the new population is most likely to be similar to the distribution of the lost population in the next period, thereby improving the matching success rate in the next period.

[0091] Specifically, based on the filled-in data of lost individuals accumulated in the current period, the target new population is searched in the population to be included, and the target new population is added to the sub-population to be included.

[0092] During implementation, a stratified variable distribution table of the cumulative lost-follow-up population can be constructed as shown in Table 1, and then this table can be used to guide the search for the target new population. That is, from the population to be included, find a population with a variable distribution as similar as possible to that of the cumulative lost-follow-up population, and add it to the sub-population to be included, thereby increasing the probability of successful matching of the lost-follow-up population in the next period.

[0093] During implementation, the matching success rate changes are monitored in each period and time period. The matching rate is an evaluation indicator of the effectiveness of strategy adjustment. The higher the value, the more effective the strategy adjustment.

[0094] In the last period j=J of the follow-up of period t, all the successfully matched substitutes in periods j=1,…,J are included in the cohort management and recorded as the total substitutes in the missing period. The total number of cohort population is N t ; and statistical analysis was performed on the stratified variables of the lost and replacement groups to assess the differences between the replacement and lost groups. It should be noted that in the difference analysis, the first type of variables of the lost individuals The true value of , we use the true value of this type of variable; and the second type of variable of the lost individuals The true value of is unknown, so we use the filled value of such variables. Specifically, the evaluation indicators for each variable include accuracy, precision, recall, and F1-score.

[0095] During implementation, non-categorical variables were stratified according to the split point, and categorical variables were naturally divided into two or more layers. The difference evaluation includes: accuracy, precision, recall, F1-score at the macro level, and accuracy, precision, recall, F1-score at the micro level. b refers to the layer, accuracy b Indicates the calculation accuracy and precision of the bth layer of the variable b Indicates the calculated precision and recall of the bth layer of the variable b represents the recall rate of the b-th layer of the variable, F1-score b represents the F1-score of the bth layer of the variable, and k(n)+1 represents the number of variable intervals or classifications, that is, the number of layers. The calculation formulas for macro and micro indicators are:

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102]

[0103]

[0104] Among them, the total number of evaluators is the total number of lost individuals who were matched with substitutes in period t.

[0105] The number of true positive individuals b That is, the stratification of lost individuals and replacement individuals is consistent, and the number of lost individuals in the same b-th stratum, the number of false positive individuals b Indicates that the stratification of lost individuals and substitute individuals is inconsistent, and the number of lost individuals whose substitute individuals are in the bth stratum; the number of true negative individuals b The number of lost individuals who are in the same stratum as the replacement individuals and are not in the bth stratum; the number of false negative individuals b It indicates that the lost individuals and the replacement individuals are not in the same classification, and the number of lost individuals in the bth stratum.

[0106]

[0107]

[0108]

[0109]

[0110] Note that the concepts of the evaluation criteria are different but after derivation the actual micro accuracy = micro precision = micro recall = micro F1-score.

[0111] The following is an example to illustrate the performance of the present invention in a real scenario.

[0112] Consider an N 0 =30 000 open cohort. T=1, J=3, 10 000 participants followed up in each period. The number of second type variables K=3, the number of first type variables L=2.

[0113] The distribution of each variable is: (ε is independent of It follows a normal distribution, and its parameters ensure and The autocorrelation coefficient between them is ρ=0.7); is a binary variable obtained by randomly drawing multiple times from a binomial distribution with parameter p=0.5. When t>0, Draw randomly from the same binomial distribution, however, There is a probability of 0.7 same; is a three-category variable obtained by random sampling multiple times from a three-category multinomial distribution with equal probability (p = 0.33). When t>0, draw randomly from the same binomial distribution, and There is a probability of 0.7 When t>0, the above steps make the same variable The time series has an autocorrelation relationship with a correlation coefficient of ρ=0.7 at adjacent time points. is an integer variable with a value range of [15,65]. is a binary variable obtained by randomly drawing multiple times from a binomial distribution with parameter p = 0.5.

[0114] The probability of loss to follow-up for an individual is: Among them, β=(0.1,-0.01,-0.35,0.45,0.019,-0.35)′. Based on this probability, each individual is randomly generated in the situation The average loss probability of the cohort population in each period is 50%.

[0115] As the follow-up of the cohort was carried out, the follow-up population was lost. We used the method of the present invention and the random inclusion method to include new populations. This example focuses on comparing the method of the present invention with the traditional method of randomly including new populations. We assume that the population to be included In order to show the distribution table of stratified variables in each period, we first conducted a single simulation to evaluate the performance of the two methods under a single simulation; then, we repeated the simulation 10 times and took the average of the 10 evaluation indicators to evaluate the two methods.

[0116] Single simulation results

[0117] t = 1 period, j = 1 time period, after two steps of S2 missing filling and S3 inclusion matching, the matching rate was 69.9%. After identifying the stratified variable distribution of the lost follow-up population that failed to match successfully, the variable stratified distribution table is shown in Table 2 below:

[0118] Table 2. Distribution of stratified variables of unmatched lost-to-follow-up population when t=1, j=1

[0119]

[0120] In the strategy adjustment step, according to the above variable stratified distribution table, a target sub-population is formed, and a new population with a corresponding distribution and number of people is found in the total population to be included, and replaced in the cohort. Furthermore, according to the stratified variable distribution table of all lost individuals when j=1, a corresponding distribution of the non-cohort population is found and added to the sub-population to be included in the next observation period.

[0121] t = 1 period, j = 2 period, after two steps of S2 missing filling and S3 inclusion matching, the matching rate was 85.1%. Compared with the first period, the matching rate was greatly improved. The distribution table of stratified variables of the unmatched lost population is shown in Table 3:

[0122] Table 3. Coarsened distribution of unmatched lost contacts when t=1, j=2

[0123]

[0124]

[0125] In the period t=1, j=3, after two steps of S2 missing filling and S3 inclusion matching, the matching rate reached 92.8%. Compared with the second period, the matching rate was greatly improved again.

[0126] Finally, we observed the follow-up effect evaluation. Variables and The true values ​​of the variables are known, so we can use the true values ​​for evaluation. To be consistent with the real world, we also use the imputed values ​​for evaluation.

[0127] Figure 3 The results of a single simulation are shown in the figure, which respectively show the effect comparison bar chart of the method of the present invention and the random inclusion method based on the real value and the filled value, multiple variables and multiple evaluation indicators in period t=1. The study shows that the method of the present invention is better in all indicators.

[0128] Figure 4 The results of multiple simulations are shown in the figure, which respectively show the effect comparison bar chart of the method of the present invention and the random inclusion method based on the real value and the filled value, multiple variables and multiple evaluation indicators in the t=1 period. The study shows that the method of the present invention is better in all indicators.

[0129] The present invention effectively solves the problem of variable differences between new participants and lost participants in open cohort studies by introducing a dynamic balance mechanism, thereby improving the external validity of cohort studies. This method can flexibly adapt to the needs of high population mobility in modern research, comprehensively reflect the overall health trends of the population, and provide more reliable data support for the study of disease mechanisms and prevention strategies. Compared with traditional closed cohort methods, the present invention can better reflect the characteristics of the target population, reduce sample bias, and improve the reliability and credibility of the conclusions. In addition, the present invention has the advantages of simple operation and wide applicability, and can be widely used in cohort studies in medical health research.

[0130] A specific embodiment of the present invention discloses an open queue construction system with a dynamic balancing mechanism, such as Figure 5 As shown, it includes the following modules:

[0131] A lost-follow-up individual determination module is used to determine the lost-follow-up individual in the current period in each period of each period of the follow-up process;

[0132] A data filling module is used to fill in the data of lost individuals based on the follow-up data of individuals in the team;

[0133] The precise matching module is used to match the missing individuals with replacement individuals in the sub-population to be included based on the filling data of the missing individuals, using coarse-grained precise matching, and adding the successfully matched replacement individuals to the cohort;

[0134] The distribution matching module is used to determine the stratified variable distribution of the unmatched lost individuals based on the filled data of the unmatched lost individuals if there are any unmatched lost individuals, and to find replacement individuals for the unmatched lost individuals in the population to be included based on the stratified variable distribution of the unmatched lost individuals to add to the queue.

[0135] Preferably, the data of lost individuals are filled based on the follow-up data of the individuals in the team, including:

[0136] The first type of variables for individuals lost to follow-up were directly imputed;

[0137] For the second type variables of the lost individuals, the propensity score matching method was used to match similar individuals to the lost individuals from the individuals in the team, and the second type variables of the lost individuals were filled based on the second type variable values ​​of the similar individuals;

[0138] Among them, the first type of variables are variables that do not change with time or that change with time but the change pattern is known, and the second type of variables are variables that change with time and the change pattern is unknown.

[0139] Preferably, the following formula is used to fill in the second type of variables of the lost individuals based on the variable values ​​corresponding to similar individuals:

[0140]

[0141] X ·k represents the kth second type variable, represents the value of the kth second type variable of the mth similar individual at the tth period, Mode(·) represents the calculation of the mode, M represents the number of similar individuals, It represents the imputed value of the kth second-type variable of the i-th individual at the t-th period, and the i-th individual is the lost individual.

[0142] Preferably, the system further comprises:

[0143] The strategy adjustment module is used to find the target new population in the population to be included based on the filling data of the lost individuals accumulated in the current period if the current period is not the last period of the current period, and add the target new population to the sub-population to be included.

[0144] The above method embodiments and system embodiments are based on the same principle, and their related parts can be used for reference, and can achieve the same technical effect. The specific implementation process refers to the above embodiment, which will not be repeated here.

[0145] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0146] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for constructing an open queue with a dynamic balancing mechanism, characterized in that: The following steps are involved: In each period of each phase of the follow-up process, the individuals lost to follow-up in the current period were identified; The data of lost individuals were imputed based on the follow-up data of individuals in the team; Based on the imputed data of lost individuals, coarse-grained precise matching is used to match the lost individuals with replacement individuals in the sub-population to be included, and the successfully matched replacement individuals are added to the cohort; If there are unmatched lost individuals, the stratification variable distribution of the unmatched lost individuals is determined based on the imputed data of the unmatched lost individuals, and replacement individuals for the unmatched lost individuals are found in the population to be included based on the stratification variable distribution of the unmatched lost individuals to join the cohort.

2. The method for constructing an open queue with a dynamic balancing mechanism according to claim 1, characterized in that: The data of lost individuals are filled based on the follow-up data of the individuals in the team, including: The first type of variables for individuals lost to follow-up were directly imputed; For the second type variables of the lost individuals, the propensity score matching method was used to match similar individuals to the lost individuals from the individuals in the team, and the second type variables of the lost individuals were filled based on the second type variable values ​​of the similar individuals; Among them, the first type of variables are variables that do not change with time or that change with time but the change pattern is known, and the second type of variables are variables that change with time and the change pattern is unknown.

3. The method for constructing an open queue with a dynamic balancing mechanism according to claim 2, characterized in that: The following formula is used to fill in the second type of variables for lost individuals based on the corresponding variable values ​​of similar individuals: X ·k represents the kth second type variable, represents the value of the kth second type variable of the mth similar individual at the tth period, Mode(·) represents the calculation of the mode, M represents the number of similar individuals, It represents the imputed value of the kth second-type variable of the i-th individual at the t-th period, and the i-th individual is the lost individual.

4. The method for constructing an open queue with a dynamic balancing mechanism according to claim 2, characterized in that: The propensity score of each individual who is lost to follow-up in period t is calculated based on the following formula, and the M individuals whose propensity scores are closest to the lost individuals are selected from the individuals in the team as similar individuals: in, represents the propensity score of the i-th individual being lost to follow-up in period t, indicates that the i-th individual was lost to follow-up in period t, represents the second type variable of the i-th individual at the t-1 period, represents the first type variable of the ith individual at the t-1th period, and P(·) represents the probability.

5. The method for constructing an open queue with a dynamic balancing mechanism according to claim 1, characterized in that: Coarse-grained precise matching is used to match replacement individuals for lost individuals in the sub-population to be included, including: Stratification was performed for each non-categorical variable at follow-up, and stratification remained unchanged for each categorical variable at follow-up; For each individual lost to follow-up, the stratum of each variable of the individual lost to follow-up was determined based on the imputed data of each variable of the individual lost to follow-up; If there is an individual in the sub-population to be included whose stratification variable distribution is consistent with that of the lost individual, the matching is successful, and an individual with a consistent stratification variable distribution is randomly selected as a replacement individual for the lost individual.

6. The method for constructing an open queue with a dynamic balancing mechanism according to claim 1, further comprising: If the current time period is not the last time period of the current period, the target new population is searched in the population to be included based on the filling data of the lost individuals accumulated in the current period, and the target new population is added to the sub-population to be included.

7. An open queue construction system with a dynamic balancing mechanism, characterized in that: Includes the following modules: A lost-follow-up individual determination module is used to determine the lost-follow-up individual in the current period in each period of each period of the follow-up process; A data filling module is used to fill in the data of lost individuals based on the follow-up data of individuals in the team; The precise matching module is used to match the missing individuals with replacement individuals in the sub-population to be included based on the filling data of the missing individuals, using coarse-grained precise matching, and adding the successfully matched replacement individuals to the cohort; The distribution matching module is used to determine the stratified variable distribution of the unmatched lost individuals based on the filled data of the unmatched lost individuals if there are any unmatched lost individuals, and to find replacement individuals for the unmatched lost individuals in the population to be included based on the stratified variable distribution of the unmatched lost individuals to add to the queue.

8. The open queue construction system with dynamic balancing mechanism according to claim 7, characterized in that: The data of lost individuals are filled based on the follow-up data of the individuals in the team, including: The first type of variables for individuals lost to follow-up were directly imputed; For the second type variables of the lost individuals, the propensity score matching method was used to match similar individuals to the lost individuals from the individuals in the team, and the second type variables of the lost individuals were filled based on the second type variable values ​​of the similar individuals; Among them, the first type of variables are variables that do not change with time or that change with time but the change pattern is known, and the second type of variables are variables that change with time and the change pattern is unknown.

9. The open queue construction system with dynamic balancing mechanism according to claim 8, characterized in that: The following formula is used to fill in the second type of variables for lost individuals based on the corresponding variable values ​​of similar individuals: X ·k represents the kth second type variable, represents the value of the kth second type variable of the mth similar individual at the tth period, Mode(·) represents the calculation of the mode, M represents the number of similar individuals, It represents the imputed value of the kth second-type variable of the i-th individual at the t-th period, and the i-th individual is the lost individual.

10. The open queue construction system with a dynamic balancing mechanism according to claim 7, further comprising: The strategy adjustment module is used to find the target new population in the population to be included based on the filling data of the lost individuals accumulated in the current period if the current period is not the last period of the current period, and add the target new population to the sub-population to be included.

Images