Construction method of myopia prediction model

Abstract

The invention discloses a method for constructing a myopia prediction model, relates to the technical field of biomedicine, and particularly aims to solve the problems that in the background technology, existing prediction models rarely comprise perinatal period factors such as fetal period and infant period factors and mother pregnancy period health conditions, and the factors are likely to be highly related to early-stage myopia incidence, so that the probability of occurrence of myopia is low. In order to solve the problem of low prediction precision caused by short-sightedness in the prior art, the invention provides a method for constructing a short-sightedness prediction model, the prediction model contains factors such as eye biological parameters, eye using behaviors, lifestyles, whether parents are short-sighted or not and whether the parents smoke or not, and also contains related factors in the perinatal period, so that short-sightedness conditions of children are accurately evaluated, and the short-sightedness prediction model can be used for predicting short-sightedness of children. The myopia tendency is expected to be found earlier, and a multi-angle theoretical basis is provided for myopia prevention and control.

Description

Technical Field

[0001] The invention belongs to the field of biomedical technology, and in particular relates to a method for constructing a myopia prediction model. Background Art

[0002] In the absence of effective intervention measures, the global prevalence of myopia is expected to reach 50% by 2050. Myopia, especially complications related to high myopia, such as myopic macular degeneration, retinal detachment, glaucoma and other vision problems, may become a serious social problem by 2050. The age of onset and duration of progression of myopia are the most important predictors of whether it will develop into high myopia in the future. Therefore, it is very necessary to detect the "signs" of myopia early and intervene in time.

[0003] The cause of myopia is still unclear. Current research reports mainly focus on influencing factors. The results show that genetic and environmental factors significantly affect the occurrence of myopia in adolescents and children, as well as the future progression of myopia and the occurrence of related complications. Outdoor activities are recognized as a protective factor against the occurrence of myopia. Risk factors include excessive eye load at close range, circadian rhythm disorders, lack of sleep, and excessive use of video devices. In the study of environmental factors, most of them focus on environmental factors in preschool or school-age children, and only a few studies focus on factors in infancy or the fetal period. Current preventive measures for myopia are mainly aimed at behavioral control of influencing factors. Therefore, it is a challenge for public health to detect signs of myopia in children as early as possible and intervene or monitor them in a timely manner.

[0004] Prediction models can use single or multiple predictors to predict the probability of an individual experiencing a certain outcome event in the future. Thanks to a large number of studies on myopia risk factors, different types of prediction models have been developed to predict the risk of myopia or high myopia in different populations. Existing myopia prediction models differ in terms of myopia definition, predictors, statistical methods, and prediction effects. In 2022, Li SM et al. used machine learning methods to construct a prediction model. The predictors included uncorrected ciliary muscle paralysis hyperopia SE, axial ratio ratio (AL / CR), K1, gender, and parental myopia. The prediction accuracy was >80%. Guo et al. used gender, SE, AL, CR, and positive relative adjustment to construct a multivariate Cox proportional hazard prediction model. The area under the receiver operating characteristic curve (ROC curve) of the validation group was 0.86. This study believes that SE after ciliary muscle paralysis is the best predictor for predicting the occurrence of myopia. At present, the predictors in this type of study mainly include SE after ciliary body paralysis, ocular biological parameters, lifestyle, parental myopia, and eye habits. Less commonly included are perinatal factors such as fetal and infant factors and maternal health status during pregnancy, which may be highly correlated with the early onset of myopia. Because premature overconsumption of hyperopia reserve is a predictor of rapid and premature onset of myopia, and the influencing factors of premature overconsumption of hyperopia reserve, in addition to personal behavior, parental factors, and lifestyle, also include pregnancy complications, history of premature birth, and breastfeeding.

[0005] Therefore, it is urgent to design a method for constructing a myopia prediction model to solve the above problems. Summary of the invention

[0006] In view of the defects of the above-mentioned prior art, the purpose of the present invention is to provide a method for constructing a myopia prediction model. In addition to incorporating factors such as the subject's eye biological parameters, eye behavior, lifestyle, whether the parents are myopic and whether the parents smoke, the prediction model also incorporates relevant factors of the perinatal period to conduct a comprehensive evaluation of the myopia situation of adolescents, in order to detect myopia tendencies earlier and provide a multi-angle theoretical basis for myopia prevention and control.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A method for constructing a myopia prediction model, characterized by comprising:

[0009] S1. Collect data related to myopia and remove incomplete data from the collected data;

[0010] S2. Assign variables to the data obtained in step S1;

[0011] S3. Filter the data after variable assignment, and divide the filtered data into a training set and a validation set;

[0012] S4. Based on the training set, a prediction model was established using multivariate binary logistic regression analysis;

[0013] S5. Use the validation set and external validation data to verify the performance of the constructed prediction model.

[0014] Preferably, in step S1, the data related to myopia include: the subject's general condition, eye habits, living habits, birth conditions, infancy conditions, mother's pregnancy conditions, parents' general condition and myopia conditions, parents' living conditions, sports ability score, naked eye vision, slit lamp examination conditions, computer optometry conditions, intraocular pressure, and eye biological parameters.

[0015] Preferably, in step S3, the screened data include: age, gender, AL, CCT, steep corneal curvature (K2), axial ratio and central anterior chamber depth, parents' myopia, whether the diet is mainly coarse grains, whether the baby is premature, mother's history of gestational hypertension, whether the parents smoke, child's personality, daily video device usage time, sports performance, father's myopia, mother's myopia, and BMI.

[0016] Preferably, in step S4, the process of establishing a prediction model using multivariate binary logistic regression analysis includes:

[0017] S401. Taking myopia as the dependent variable and the index screened in step S3 as the independent variable, a binary logistic regression analysis is performed;

[0018] S402. Set P < 0.05 and obtain data for establishing a prediction model;

[0019] S403. Based on the data for establishing the prediction model, the equation is transformed according to the binary logistic regression equation to obtain the prediction model expression.

[0020] Preferably, in step S402, the data for establishing the prediction model include: age, K2, AL, whether the parents are myopic, AL / CR, and the mother's history of gestational hypertension.

[0021] Preferably, the prediction model expression is:

[0022] P=1 / [1+e-(-168.2+0.2589*X1+1.698*X2+4.215*X3+0.8789*X4-3.126*X5-4.12*X6 / -5.699*X7 / -6.573*X8 / -7.876*X9)]

[0023] In the formula, X1 is age; X2 is corneal curvature K2; X3 is axial length; X4 is myopia of both parents; X5 is the history of gestational hypertension in the mother; X6 is axial ratio 2.85-2.9; X7 is axial ratio 2.9-2.95; X8 is axial ratio 2.95-3.00; X9 is axial ratio ≥3.00, P is the probability of disease, and e is the natural logarithm.

[0024] Preferably, in step S5, the external verification data is from a source different from the data in step S1.

[0025] The beneficial effects of the present invention are as follows: the present invention discloses a method for constructing a myopia prediction model. Compared with the prior art, the present invention has the following improvements:

[0026] The present invention incorporates perinatal factors, such as pregnancy complications, premature birth, etc., which are factors that cause the hyperopia reserve to be consumed too quickly, into the prediction index, and the hyperopia reserve is consumed too quickly, which is an important indicator for predicting myopia. Therefore, the present invention incorporates perinatal factors into the construction of the myopia prediction model, and can predict the myopia trend more comprehensively and accurately, in order to provide a theoretical basis for earlier detection of myopia tendency. In addition, the present invention verifies the model outside the population, confirming the accuracy and practicality of the prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flowchart of a method for constructing a myopia prediction model according to the present invention;

[0028] Figure 2 It is the ROC curve diagram in the internal validation set of the present invention;

[0029] Figure 3 This is a calibration test diagram for the internal verification of the present invention;

[0030] Figure 4 It is the ROC curve diagram of the present invention in the external validation set;

[0031] Figure 5 This is a calibration test diagram for the present invention in the external validation set. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0033] Example:

[0034] See attached Figure 1 As shown, the method for constructing a myopia prediction model includes:

[0035] S1. Collect data related to myopia and remove incomplete data from the collected data.

[0036] The data in this step are obtained by surveying and counting subjects from the following sources: Source 1. Students in grades 1-3 who underwent eye health examinations at a primary school in Yanqing District, Beijing from October 1, 2022 to October 31, 2022; Source 2. Students who underwent myopia examinations at the outpatient clinic of Yanqing District Hospital, Beijing from December 1, 2022 to January 31, 2023; Source 3. Students in grades 2-5 who underwent eye health examinations at another primary school in Yanqing District, Beijing from November 1, 2023 to December 31, 2023; Source 4. Students who underwent myopia examinations at a hospital in downtown Beijing from November 1, 2023 to December 31, 2023;

[0037] The survey content includes: the general condition of the subjects, eye habits, living habits, birth conditions, infancy conditions, mother's pregnancy conditions, parents' general condition and myopia conditions, parents' living habits, and eye data. Specific:

[0038] (1) The general information of the subjects included: age, gender, height, weight, personality, whether they suffered from allergic rhinitis, and whether they were currently myopic.

[0039] (2) Subjects’ eye habits included: daily screen time;

[0040] (3) Subjects’ living habits included: whether the diet was mainly meat-based, weekly consumption of coarse grains, daily outdoor activity time, and daily sleep time.

[0041] (4) The birth conditions of the subjects included: birth weight (kg), whether they were born prematurely, whether the umbilical cord was wrapped around the neck during the fetal period, and the mode of delivery.

[0042] (5) Infant conditions: feeding conditions (breast milk, formula, mixed feeding), history of incubator treatment after birth, and history of blue light treatment after birth.

[0043] (6) Maternal pregnancy status: whether the mother has gestational diabetes, gestational hypertension, anemia during pregnancy, history of viral infection during pregnancy (cold, herpes, positive viral infection in TORCH examination).

[0044] (7) General condition of parents and their myopia: whether the father is myopic, whether the mother is myopic, whether there is a family history of hereditary myopia, the father’s educational background, and the mother’s educational background.

[0045] (8) Parents’ living conditions: whether parents smoke.

[0046] (9) Subjects’ physical ability scores:

[0047] (10) Naked eye vision test: Perform vision test in accordance with the provisions of the Standard Naked Eye Visual Acuity Logarithmic Chart. Method: In a dark room, use fluorescent lighting, and the illumination of the visual acuity chart is 300-500LUX, ensuring that the brightness of the visual acuity chart is higher than that of the test environment. Ask the test subject to take off his frame glasses and stand 5 meters away from the visual acuity chart, with his back to the light source and his body upright. Keep his line of sight flush with the 5.0 sight mark on the visual acuity chart. Check both eyes independently, first check the right eye and then the left eye. The test subject covers the eye opposite to the test eye and tests it. Ask the test subject not to squint or oppress the covered eye. Test from the first row of sight marks from top to bottom, and the test subject should point out the direction of the sight marks within 3 seconds. The visual acuity corresponding to the last row of sight marks that can be pointed out correctly is the naked eye visual acuity of the corresponding eye. The result is recorded as accurate if all sight marks in a row can be identified. If the test subject's vision is estimated to be good, it is not necessary to start from the largest row of sight marks, and it can be started from a smaller row as appropriate.

[0048] (11) Slit lamp examination: Assess the health of the eyelids, conjunctiva, cornea, anterior chamber, and lens, and exclude anterior segment-related diseases.

[0049] (12) Computerized optometry: A computerized optometry instrument (KR-1, Topcon, Japan) was used to examine the refraction, and the average value was taken after three measurements.

[0050] (13) Intraocular pressure measurement: Intraocular pressure was measured using a non-contact tonometer (Topcon, Japan) under fluorescent lighting, and the average value was obtained by three measurements.

[0051] (14) Measurement of ocular biological parameters: The ocular biological parameters were measured using an ophthalmic optical biometer LS900 (Haag-Streit Koeniz, Switzerland). The measurements were repeated three times and the average value was taken. The AL, ACD, CCT, lens thickness, corneal curvature, and white-to-white were recorded for the right and left eyes respectively.

[0052] (15) After the survey, 5% of the subjects were randomly selected to remeasure the relevant data and evaluate their accuracy.

[0053] The exclusion criteria for the subjects were as follows: (1) those with metabolic or congenital systemic diseases; (2) those with congenital cataracts, glaucoma, uveitis or corneal lesions and other eye diseases; (3) those with hyperopia (>+2.00D), strabismus or amblyopia; (4) those with a history of intraocular surgery; (5) those with a history of orthokeratology lenses or low-concentration atropine; (6) those with a family history of hereditary eye diseases; (7) those whose guardians refused to provide informed consent. The final number of subjects was 1,304, aged between 4 and 12 years old, with an average age of (7.85 +1.58) years old, 306 people are myopic, 998 people are not myopic, and the myopia rate is 23.5%. Among them, there are 679 males, 158 of whom are myopic, and the myopia rate is 23.2%; there are 625 females, 142 of whom are myopic, and the myopia rate is 22.7%.

[0054] The present invention also analyzes the effects of less reported factors, such as history of premature birth, history of gestational hypertension during pregnancy, history of gestational diabetes during pregnancy, history of incubator treatment after birth, and history of blue light treatment after birth on the onset of myopia and eye biological parameters. Specifically,

[0055] In the premature population, age is a protective factor for myopia, while height and father's myopia are independent risk factors for myopia. Older age is associated with thicker CCT, and the time spent outdoors for more than 2 hours per day is associated with thinner CCT.

[0056] In the non-premature population, independent risk factors for myopia include paternal myopia, maternal myopia, age, ACD, and AL; independent risk factors for AL include height, paternal myopia, maternal myopia, and daily outdoor activity time >2 hours. Maternal history of viral infection during pregnancy is a protective factor. Subjects who use video equipment for >2 hours a day and whose fathers smoke have thinner CCT. Subjects' age*weight are independent risk factors for CCT.

[0057] In boys, increased birth weight was associated with a flatter K1; independent risk factors for AL included height, paternal myopia, and maternal myopia; daily outdoor activity time >2 hours and a history of maternal viral infection during pregnancy were protective factors.

[0058] Among the factors affecting K1 in girls, height and maternal history of gestational diabetes were protective factors, while history of allergic rhinitis was a risk factor. Independent risk factors for AL included height and paternal myopia.

[0059] Among people with a history of maternal viral infection during pregnancy, the risk factors for K2 include gender, maternal myopia, and maternal education level. That is, girls, maternal myopia, and subjects with a higher maternal education level have higher K2 values, and those who deliver naturally have higher K2 values. Among people without a history of maternal viral infection, female subjects have higher K2 values. Subjects with higher birth weight, maternal myopia, and maternal education level have lower K2 values.

[0060] Risk factors associated with ACD include height, age, father's myopia and mother's education level. Males have deeper ACD.

[0061] Before the age of 12, a history of premature birth is an influencing factor for myopia and CCT, and myopia in premature infants may not be related to the growth of AL. A history of viral infection during pregnancy is a common influencing factor for corneal curvature and AL. Outdoor activities for more than 2 hours a day may only be effective for boys' axial length control, and the causes and mechanisms of myopia in boys and girls may be different. Therefore, in the study and prevention of myopia, the population should be subdivided and precise prevention and control should be carried out.

[0062] S2. In view of the influence of pregnancy factors on eye development, in order to facilitate statistical analysis, the data obtained in step S1 are assigned variables; the details are as follows:

[0063] (1) BMI: BMI was converted into a categorical variable based on the interquartile range: 0: less than 14.67; 1: 14.67-16.57; 2: 15.57-19.74; 3: greater than 19.74.

[0064] (2) Myopia of parents: 0: neither parent has myopia; 1: one parent has myopia; 2: both parents have myopia

[0065] (3) AL / CR: According to the literature, it is divided into 6 levels: 0: less than 2.8; 1: 2.8-2.85; 2: 2.85-2.9; 3: 2.9-2.95; 4: 2.95-3; 5: greater than 3.

[0066] (4) Video equipment usage time: 0: less than 1 hour; 1: 1 to 2 hours; 2: more than 2 hours.

[0067] (5) The general information of the subjects included: gender (1: male, 2: female); personality (1: introvert, 2: extrovert); whether they suffered from allergic rhinitis (1: yes, 2: no); and whether they were currently myopic (0: no; 1: yes).

[0068] (6) Subjects’ eye habits included: daily screen usage time (0: <2 hours, 1: ≥2 hours).

[0069] (7) Subjects’ living habits included: whether the diet was mainly meat-based (1: yes, 2: no), weekly coarse grain consumption (0: every day; 1: 2-3 days / week; 2: occasionally; 3: never), daily outdoor activity time (0: ≥2 hours; 1: <2 hours), and daily sleep time (0: 8-10 hours; 1: 6-8 hours; 2: less than 6 hours).

[0070] (8) The birth data of the subjects included: birth weight (kg), whether they were born prematurely ((1: yes, 2: no), whether the umbilical cord was wrapped around the neck during the fetal period ((1: yes, 2: no), and mode of delivery (1: normal delivery; 2: cesarean section).

[0071] (9) Infant status: feeding status (1. breast milk; 2. formula milk; 3. mixed feeding), whether there is a history of incubator treatment after birth (1: yes, 2: no), whether there is a history of blue light treatment after birth (1: yes, 2: no).

[0072] (10) Maternal pregnancy status: whether the mother had gestational diabetes (1: yes, 2: no), whether the mother had gestational hypertension (1: yes, 2: no), whether the mother had anemia during pregnancy (1: yes, 2: no), whether the mother had a history of viral infection during pregnancy (cold, herpes, positive viral infection in TORCH examination, 1: yes, 2: no).

[0073] (11) General and myopia status of parents: whether the father is myopic (1: no; 2: yes), whether the mother is myopic (1: no; 2: yes), whether there is a family history of myopia (1: yes, 2: no), father's education level (1. high school or below; 2. college or above), mother's education level (1. high school or below; 2. college or above).

[0074] (12) Parents’ living conditions: whether parents smoke (0: neither parent smokes; 1: one parent smokes; 2: both parents smoke).

[0075] (13) Subjects’ physical ability score: 0: 80-100 points; 1: 60-80 points; 2: less than 60 points.

[0076] S3. Filter the data after variable assignment, and divide the filtered data into a training set and a validation set in a ratio of 8:2; the training set is used to establish the prediction model, and the validation set is used to verify the model performance.

[0077] In this step, the data after variable assignment is preliminarily screened by LASSO regression to obtain the selected indicators. The data after variable assignment include: gender, age, parents' myopia, whether the diet is mainly coarse grains, whether breastfeeding, whether there is a family history of myopia, whether it is a premature baby, whether the diet is mainly vegetables, delivery method, whether there is a history of umbilical cord around the neck during the fetal period, whether the mother has a history of gestational diabetes, whether the mother has a history of gestational hypertension, whether the parents smoke, whether there is a history of blue light treatment after birth, whether there is a history of incubator treatment after birth, mother's history of anemia during pregnancy, mother's history of infection during pregnancy, history of allergic rhinitis, child's personality, daily video equipment use time, daily outdoor activity time, sports performance, Father's education level, mother's education level, BMI, AL, CCT, steep corneal curvature (K2), axial ratio ratio and central anterior chamber depth were screened using LASSO regression. Specifically, when the regression penalty term was set to λ=0.01312276, the selected indicators included: age, gender, AL, CCT, steep corneal curvature (K2), axial ratio ratio and central anterior chamber depth, parents' myopia, whether the diet was mainly coarse grains, whether the baby was a premature baby, mother's history of gestational hypertension, whether the parents smoked, the child's personality, daily video device usage time, sports performance, father's myopia, mother's myopia, and BMI.

[0078] S4. Based on the training set, a prediction model was established using multivariate binary logistic regression analysis:

[0079] Specifically, whether myopia is a dependent variable and the index screened in step S3 are used as independent variables to perform binary logistic regression analysis, and the data in the training set with p<0.05 (p is a commonly used variable in statistics, and p<0.05 is generally recognized as a meaningful index) are included in the prediction model. The data included in the prediction model include: age, K2, AL, parental myopia, AL / CR and maternal history of gestational hypertension; according to the binary logistic regression equation, P(y=1|x)=π(x)=1 / l+e -g(x) , transform the equation and get the prediction model expression as follows:

[0080] P=1 / [1+e-(-168.2+0.2589*X1+1.698*X2+4.215*X3+0.8789*X4-3.126*X5-4.12*X6 / -5.699*X7 / -6.573*X8 / -7.876*X9)]

[0081] In the formula, X1 is age; X2 is corneal curvature K2; X3 is axial length; X4 is myopia of both parents; X5 is the history of gestational hypertension in the mother; X6 is axial ratio 2.85-2.9; X7 is axial ratio 2.9-2.95; X8 is axial ratio 2.95-3; X9 is axial ratio ≥3, P is the probability of disease occurrence, and e is the natural logarithm.

[0082] S5. Based on the validation set and external validation data, the prediction performance of the model is verified respectively;

[0083] The validation process is divided into two parts: one is internal validation, that is, validation of the internal validation data set; the other is external validation by introducing external data (student data from another school different from the school where the model was developed and physical examination data from two hospitals) to calculate the discrimination and calibration of the prediction model in different data in order to evaluate the extrapolation of the model.

[0084] The results are as follows: Figure 2-3 As shown in Figure 2, the sensitivity of the model is 86.5%, the specificity is 71.4%, and the accuracy (AUC) is 83.7%. The calibration test uses the Hosmer-Lemeshow goodness-of-fit test, and the results show that the P value is 0.5555, which means that the model is well calibrated. Figure 4-5 As shown, the model had a sensitivity of 87.7%, a specificity of 78.4%, and an accuracy (AUC) of 88.6%. The calibration test was performed using the Hosmer-Lemeshow goodness-of-fit test, and the results showed that the P value was 0.2213, indicating that the model had good calibration. The AUC of the model was greater than 80% in both internal and external validation, and the accuracy and calibration were good, making it suitable for promotion.

[0085] Abbreviations / symbols

[0086]

[0087] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for constructing a myopia prediction model, characterized in that: include: S1. Collect data related to myopia and remove incomplete data from the collected data; S2. Assign variables to the data obtained in step S1; S3. Filter the data after variable assignment, and divide the filtered data into a training set and a validation set; S4. Based on the training set, a prediction model was established using multivariate binary logistic regression analysis; S5. Use the validation set and external validation data to verify the performance of the constructed prediction model.

2. The method for constructing a myopia prediction model according to claim 1, characterized in that: In step S1, the data related to myopia include: the subject's general condition, eye habits, living habits, birth conditions, infancy conditions, mother's pregnancy conditions, parents' general condition and myopia conditions, parents' living conditions, sports ability score, naked eye vision, slit lamp examination conditions, computer optometry conditions, intraocular pressure, and eye biological parameters.

3. The method for constructing a myopia prediction model according to claim 1, characterized in that: In step S3, the screened data include: age, gender, AL, CCT, steep corneal curvature (K2), axial ratio and central anterior chamber depth, parents' myopia, whether the diet is mainly coarse grains, whether the baby is premature, mother's history of gestational hypertension, whether the parents smoke, the child's personality, daily video device usage time, sports performance, father's myopia, mother's myopia, and BMI.

4. The method for constructing a myopia prediction model according to claim 1, characterized in that: In step S4, the process of establishing a prediction model using multivariate binary logistic regression analysis includes: S401. Taking myopia as the dependent variable and the index screened in step S3 as the independent variable, a binary logistic regression analysis is performed; S402. Set P < 0.05 and obtain data for establishing a prediction model; S403. Based on the data for establishing the prediction model, the equation is transformed according to the binary logistic regression equation to obtain the prediction model expression.

5. The method for constructing a myopia prediction model according to claim 4, characterized in that: In step S402, the data for establishing the prediction model include: age, K2, AL, whether the parents are myopic, AL / CR, and the mother's history of gestational hypertension.

6. The method for constructing a myopia prediction model according to claim 4, characterized in that: In step S403, the prediction model expression is: P=1 / [1+e-(-168.2+0.2589*X1+1.698*X2+4.215*X3+0.8789*X4-3.126*X5-4.12*X6 / -5.699*X7 / -6.573*X8 / -7.876*X9)] In the formula, X1, age; X2, corneal curvature K2; X3, axial length; X4, both parents are myopic; X5, mother has a history of hypertension during pregnancy; X6, axial rate ratio 2.85-2.9; X7, axial rate ratio 2.9-2.95; X8, axial rate ratio 2.95-3.00; X9, axial rate ratio ≥3.00, P is the probability of disease, and e is the natural logarithm.

7. The method for constructing a myopia prediction model according to claim 1, characterized in that: In step S5, the external verification data is different from the data source in step S1.

Images