Chronic painful temporomandibular joint disorder early risk prediction model establishment method, model, system, prediction method and application

Abstract

The invention discloses a method for establishing an early risk prediction model for chronic painful temporomandibular joint disorder, and the method comprises the following steps: 1, collecting the biological, psychological and social information of a subject, and obtaining 18 potential risk factors; 2, screening potential risk factors by using single-factor analysis and multi-factor logistic regression; and step 3, establishing and verifying a risk prediction model based on the screened risk factors. The invention further discloses a chronic painful temporomandibular joint disorder early-stage risk prediction model established according to the method, and an early-stage risk prediction system established on the basis of the chronic painful temporomandibular joint disorder early-stage risk prediction model. Meanwhile, the invention further discloses an early risk prediction method for the chronic painful temporomandibular joint disorder disease, and the method has wide application value.

Description

Technical Field

[0001] The present invention belongs to the technical field of risk prediction, and relates to a method for establishing an early risk prediction model for chronic pain temporomandibular joint disorder, a model, a system, a prediction method and an application thereof. Background Art

[0002] Temporomandibular joint disorder (TMD) is the most common oro-facial pain disease. After some TMD patients receive non-invasive treatment, the pain still persists and develops into chronic pain. In addition to treatment methods, there are still important risk factors affecting the prognosis of the disease.

[0003] Currently, there are still certain limitations in the construction and presentation techniques of risk prediction models for chronic pain temporomandibular joint disorder after non-invasive treatment. One of the risk prediction models for persistent temporomandibular joint disorder pain constructed in the prior art found that pain in other parts of the body, depression, oral behavior and mandibular function were significantly correlated with pain persistence, but the sample size included was small, only 63 subjects were included, making it difficult to test the prediction ability of the model. The prior art has also confirmed the influence of psychological factors on long-term temporomandibular joint disorder pain, but the risk factors considered are not comprehensive, only age, gender and psychological factors are included, affecting the reliability of the model. At the same time, the prediction models constructed in the prior art can only calculate the occurrence risk of chronic pain through complex formulas, and the calculation method is complex and time-consuming, which limits the clinical application and promotion.

[0004] In addition, 1. The sample size included in the existing model is too small (63 cases), making it difficult to test the prediction ability of the model; 2. The risk factors included in the existing model are not comprehensive (age, gender, psychological factors), affecting the reliability of the model; 3. The existing model is usually presented by complex formulas, which limits its applicability, promotion and readability. Summary of the Invention

[0005] In order to solve the deficiencies of the prior art, the purpose of the present invention is to provide a method for establishing an early risk prediction model for chronic pain temporomandibular joint disorder, a model, a system, a prediction method and an application thereof.

[0006] The present invention provides an early risk prediction model system, method and application for chronic pain temporomandibular joint disorder. A risk prediction model is a statistical tool for estimating the probability of a specific event occurring. The present invention intends to construct a risk prediction model for chronic pain temporomandibular joint disorder after non-invasive treatment through risk factor research. On this basis, a prediction system is established and a corresponding mini-program is developed, which can realize the early risk prediction of chronic pain temporomandibular joint disorder. Patients with temporomandibular joint disorder fill in relevant questionnaires on the mini-program to quantify risk factors, and can directly calculate the risk level of an individual suffering from temporomandibular joint disorder pain, which helps to identify high-risk populations at an early stage.

[0007] The present invention quantifies risk factors based on the sociodemographic, behavioral, psychological and disease-related information of TMD patients, and obtains the risk level of an individual suffering from chronic TMD pain in the form of calculating a total score, helping clinicians to identify and screen high-risk populations at an early stage, and intervening in risk factors at an early stage to achieve "early detection and early intervention" of the disease.

[0008] Specifically, aiming at problems such as too small sample size included in the prior art, insufficiently comprehensive consideration of risk factors, and complex presentation methods, the present invention, based on previous literature searches, selected 18 potential risk factors among many biological, psychological and social factors, including age, gender, body mass index (BMI), ethnicity, place of residence, marital status, oral behavior, somatization, depression, anxiety, sleep quality, diagnostic classification, disease course, pain intensity, orthodontic treatment history, previous medication history, tooth loss, treatment method, etc. 488 subjects were included, and the above 18 potential risk factors were collected on the day of the benchmark test. All subjects received non-invasive treatment, and 12 months after baseline, they were asked by telephone follow-up whether they had chronic pain. Using the above 18 potential risk factors as independent variables and whether chronic pain occurred as the dependent variable, risk factors related to the occurrence of chronic pain TMD were finally screened out through univariate and multivariate Logistic regression analyses, a risk prediction model was established and presented in the mini-program, making the use of the risk prediction model in the present invention simple, easy to read and convenient for popularization. The discrimination, calibration and clinical practicability of the model were evaluated through the C-index, calibration plot and decision curve analysis. The Bootstrap resampling method was used to internally validate the model, and the C-index was repeatedly tested to ensure the prediction performance and reliability of the model.

[0009] The present invention provides a method for establishing an early risk prediction model for chronic pain temporomandibular joint disorder, and the method includes the following steps: Step 1: Collect the biological, psychological and social information of the subjects, and screen and obtain 18 potential risk factors; Step 2: Screen potential risk factors using univariate analysis and multivariate logistic regression; Step 3: Establish and validate a risk prediction model based on the screened risk factors.

[0010] In Step 1, the 18 potential risk factors cover multiple aspects such as biology, psychology, and society, including: (1) Age: Divided into 18 - 30 years old, 31 - 40 years old, 41 - 50 years old, 51 - 60 years old, and 61 - 70 years old according to each age group; (2) Gender: Divided into male and female; (3) Body mass index (BMI): According to the adult body weight measurement guidelines issued in 2013, the patient's weight status is divided into four grades: underweight (BMI < 18.5 kg / m 2 ), normal (18.5 - 24.0 kg / m 2 ), overweight (24.1 - 28.0 kg / m 2 ), and obese (≥ 28.0 kg / m 2 ); (4) Ethnicity: Divided into Han and others; (5) Place of residence: Divided into urban and rural; (6) Marital status: Divided into married and unmarried; (7) Oral behavior: Divided into 2 groups according to the score of the Oral Behaviour Checklist (OBC). An OBC score of 0 - 24 points indicates a low frequency of oral behavior; a score of ≥ 25 points indicates a high frequency of oral behavior; (8) Somatization: Divided into 4 groups according to the score of the Patient Health Questionnaire - 15 (PHQ - 15). A PHQ - 15 score of 0 - 4 points indicates no somatization; 5 - 9 points indicates mild somatization; 10 - 14 points indicates moderate somatization; ≥ 15 points indicates severe somatization; (9) Depression: Divided into 4 groups according to the score of the Patient Health Questionnaire - 9 (PHQ - 9). A PHQ - 9 score of 0 - 4 points indicates no depression; 5 - 9 points indicates mild depression; 10 - 14 points indicates moderate depression; ≥ 15 points indicates severe depression; (10) Anxiety: Divided into 4 groups according to the score of the Generalized Anxiety Disorder - 7 (GAD - 7). A GAD - 7 score of 0 - 4 points indicates no anxiety; 5 - 9 points indicates mild anxiety; 10 - 14 points indicates moderate anxiety; ≥ 15 points indicates severe anxiety; (11) Sleep quality: Divided into two groups according to the score of Pittsburgh Sleep Quality Index (PSQI). A PSQI score of 0 - 5 indicates good sleep quality; a score > 5 indicates poor sleep quality. (12) Diagnostic classification: Divided into three groups: joint diseases, painful diseases, and mixed diseases. (13) Course of disease: Divided into < 1 month, 1 - 3 months, and > 3 months. (14) Pain intensity: According to the Numerical Rating Scale (NRS), an NRS score of 0 - 3 indicates no / mild pain, and a score > 3 indicates moderate / severe pain. (15) History of orthodontic treatment: Divided into having and not having. (16) History of previous medication: Divided into having and not having. (17) Tooth loss: Divided into having and not having. (18) Treatment methods: Divided into six groups: health education, drug treatment, physical factor treatment, drug treatment + physical factor treatment, physical factor + manual treatment, drug + physical factor + manual treatment.

[0011] In step two, the univariate analysis in the present invention preliminarily screens potential risk factors by comparing the result changes caused by the change of any one of the 18 potential risk factors. Specifically, first, through univariate analysis, the result differences caused by the change of any one of the 18 potential risk factors, namely age, gender, BMI, ethnicity, place of residence, marital status, oral behavior, somatization, depression, anxiety, sleep quality, diagnostic classification, course of disease, pain intensity, history of orthodontic treatment, history of previous medication, tooth loss, and treatment methods, are compared respectively. The risk factors initially obtained to be related to the occurrence of chronic pain TMD include oral behavior (OBC score), somatization (PHQ - 15 score), depression (PHQ - 9 score), anxiety (GAD - 7 score), sleep quality (PSQI score), diagnostic classification, and pain intensity (all P < 0.001).

[0012] The multivariate Logistic regression analysis in the present invention is a statistical method for studying the influence of multiple independent variables on the result variable. Its basic principle is to model the response variable and consider the combined effect of multiple factors on the result. In the analysis process, the maximum likelihood method is usually used to estimate the regression coefficients, and the model is used to predict the probability of the result.

[0013] The mathematical model of the multivariate Logistic regression analysis is expressed as:

[0014] Among them, P represents the probability of chronic temporomandibular disorder pain when a given combination of risk factors is present; x i represents different risk factors; β 0 represents the baseline log odds when all risk factors are 0; β j represents the regression coefficient of different risk factors; The parameters of the entire model are estimated by the maximum likelihood method. The main concepts in the model are as follows: (1) P / 1 - P: It is called the odds or advantage. In(P / 1 - P)=logit(P) is called the logarithm of the odds. A large number of practices have proved that logit(P) has a linear relationship with quantitative independent variables.

[0015] (2) OR value (Odds Ratio): Also known as the odds ratio, it mainly refers to the ratio P / 1 - P in the case group divided by the ratio P / 1 - P in the control group, which is a commonly used indicator in epidemiology and medical research.

[0016] (3) Partial regression coefficient β j (j = 1, 2...., m): It represents the change amount of Logit(P) when the independent variable changes by one unit under the condition that other conditions remain unchanged. If the regression coefficient is positive, it indicates a positive correlation between the independent variable and the dependent variable; if it is negative, it indicates a negative correlation between the independent variable and the dependent variable.

[0017] (4) The relationship between the regression coefficient and the OR value: The regression coefficient mainly interprets the significance of the independent variable and the positive and negative directions of its influence on the dependent variable. The OR value is used to measure the degree of the effect of the independent variable on the dependent variable, and the OR value is equal to the natural logarithm value of the regression coefficient.

[0018] In the present invention, the risk factors with significant differences preliminarily screened out from the above single factors are used as independent variables, and whether chronic TMD pain occurs is used as the dependent variable for multivariate Logistic regression analysis. The final risk factors screened out include: PHQ - 15 score (somatization), GAD - 7 score (anxiety), PSQI score (sleep quality), and pain intensity.

[0019] Based on the Logistic mathematical model, the formula for calculating the odds of chronic pain occurring after non - invasive treatment is expressed as follows: Odds = EXP[A + 0.690×B + 1.472×C + 3.508×D + E + 1.094×F + 1.707×G + 2.812×H + 0.815×I + 0.775×J - 5.495]; Among them, A represents the assignment for a PHQ-15 score of 0-4 points, B represents the assignment for a PHQ-15 score of 5-9 points, C represents the assignment for a PHQ-15 score of 10-14 points, D represents the assignment for a PHQ-15 score of ≥15 points, E represents the assignment for a GAD-7 score of 0-4 points, F represents the assignment for a GAD-7 score of 5-9 points, G represents the assignment for a GAD-7 score of 10-14 points, H represents the assignment for a GAD-7 score of ≥15 points, I represents the assignment for a PSQI score, and J represents the assignment for pain intensity.

[0020] The incidence of chronic pain after non-invasive treatment P = Odds / (1 + Odds) × 100%.

[0021] Assignment method: When the PHQ-15 score is 0-4 points, the assignment is 0; when the PHQ-15 score is 5-9 points, the assignment is 1; when the PHQ-15 score is 10-14 points, the assignment is 2; when the PHQ-15 score is ≥15 points, the assignment is 3; when the GAD-7 score is 0-4 points, the assignment is 0; when the GAD-7 score is 5-9 points, the assignment is 1; when the GAD-7 score is 10-14 points, the assignment is 2; when the GAD-7 score is ≥15 points, the assignment is 3; when the PSQI score is 0-5 points, the assignment is 0; when the PSQI score is >5 points, the assignment is 1; when the NRS score of the baseline pain intensity is 0-3 points, the assignment is 0; when the NRS score of the baseline pain intensity is >3 points, the assignment is 1.

[0022] In step three, based on the above 4 risk factors related to chronic pain TMD screened by Logistic regression analysis: PHQ-15 score (somatization), GAD-7 score (anxiety), PSQI score (sleep quality), and pain intensity, the present invention established a risk prediction model for chronic pain TMD after non-invasive treatment.

[0023] The present invention evaluates the discrimination, calibration, and clinical practicability of the risk prediction model through the C-index, calibration curve, and DCA decision curve; uses the Bootstrap method to perform internal validation on the model; and visualizes the risk prediction model through the nomogram method.

[0024] Based on the above construction method, the present invention constructed a risk prediction model for early risk prediction of chronic pain temporomandibular joint disorders.

[0025] The present invention also provides an early risk prediction system for chronic pain temporomandibular joint disorders. The risk prediction system includes the above risk prediction model; the risk prediction system can be loaded and embedded in a small program on a mobile phone, including an input module, a conversion module, and an output module. The input module is used to fill in the PHQ-15, PSQI, and GAD-7 self-assessment questionnaires, evaluate the pain intensity, and calculate the questionnaire scores; The conversion module is used to convert the calculated questionnaire scores in the risk prediction model according to the assignment method, and calculate the risk probability of an individual developing chronic TMD pain based on the Logistic regression analysis mathematical model; The output module is used to convert the calculated risk level into a high-risk or low-risk output result for output.

[0026] The present invention also provides an early risk prediction method for chronic pain temporomandibular joint disorder, and the method includes the following steps: Step 1: Input the PHQ-15 score, GAD-7 score, PSQI score, and pain score of the subject to be predicted into the above-mentioned risk prediction system; Step 2: According to the scoring method and the Logistic regression analysis calculation formula, the risk prediction system processes and converts the original data to obtain the risk probability of the subject developing chronic pain; Step 3: Conduct high and low risk prediction classification, and output the high and low risk results and the nomogram risk prediction model results.

[0027] The beneficial effects of the present invention include: Existing technologies have problems such as too small a sample size (63 cases) included, not comprehensive enough risk factors considered (age, gender, psychological factors), and complex presentation methods (presented by formulas). In view of the above deficiencies, the present invention incorporates a sufficient sample size (488 cases) according to the Logistic regression analysis sample size calculation method; selects 18 candidate risk factors based on previous literature searches; presents and applies the model in the form of a small program, and verifies the reliability of the model through various evaluation methods.

[0028] In summary, the present invention solves many deficiencies of the existing technologies and provides a convenient, accurate, and visual prediction tool for the occurrence of chronic pain in patients with temporomandibular joint disorder.

[0029] The present invention incorporates a sufficient sample size (488 cases) according to the Logistic regression analysis sample size calculation method, and tests the prediction ability of the model through the C-index, calibration graph, decision analysis curve, and internal validation method based on the Bootstrap resampling method; Based on previous literature searches, the present invention comprehensively incorporates sociodemographic, behavioral, psychological, and disease-related information, a total of 18 potential risk factors, to ensure the reliability of the model; The present invention presents the model in the form of a nomogram. The nomogram is based on a multi-factor Logistic regression analysis, integrating multiple independent predictors. The predictors are plotted on the same plane in proportion using scale lines, and the scores corresponding to each predictor can be obtained. The scores are added up to get the total score, which is the occurrence probability of the outcome event. Compared with complex formulas, the nomogram risk prediction model has the advantages of simplicity, intuitiveness, and visualization, increasing the readability of the results and being more suitable for clinical promotion.

[0030] The present invention presents the model and establishes a model prediction system in the form of a mini-program. Patients fill out a questionnaire on the mini-program, calculate the questionnaire score, and based on the risk prediction model, the risk level of an individual developing chronic pain can be obtained. This prediction system has the advantages of simplicity, intuitiveness, and visualization, which is conducive to increasing the readability of the results and the clinical promotability of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 It is a schematic diagram of the risk prediction nomogram model of the present invention.

[0033] Figure 2 It is an ROC curve diagram of the risk prediction model of the present invention.

[0034] Figure 3 It is a calibration diagram of the risk prediction model of the present invention.

[0035] Figure 4 It is a clinical decision curve diagram of the risk prediction model of the present invention.

[0036] Figure 5 It is a schematic diagram of the architecture of the risk prediction system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] Combined with the following specific embodiments and drawings, the present invention will be further described in detail. The processes, conditions, experimental methods, etc. for implementing the present invention, except for the specifically mentioned content below, are all common knowledge and well-known common sense in the art, and the present invention has no special limiting content.

[0038] The present invention passes through Figure 1The four independent risk factors shown: the PHQ-15 score (somatic symptom score), the GAD-7 score (anxiety score), the PSQI score (sleep quality score), and the baseline pain intensity are put into the input module of the mini-program ( Figure 5 ), and through the conversion module, according to the assignment method, the original numerical values are converted into assigned numerical values. Based on the Logistic regression analysis calculation formula, the risk probability of developing chronic TMD pain is calculated. A risk of ≥50% is set as high risk, and <50% is set as low risk. The results (high risk / low risk) and the nomogram risk prediction model ( Figure 1 ) are output through the output module.

[0039] Example 1

[0040] In this example, an early risk prediction model for chronic pain-related temporomandibular disorders was first constructed based on the following method, and the method includes the following steps: Step 1: Collect the biological, psychological, and social information of the subjects, and screen and obtain 18 potential risk factors; Step 2: Use univariate analysis and multivariate logistic regression to screen the potential risk factors; Step 3: Establish and validate a risk prediction model based on the screened risk factors.

[0041] In step 1, the potential risk factors include age, gender, BMI, ethnicity, place of residence, marital status, oral behavior, somatic symptoms, depression, anxiety, sleep quality, diagnostic classification, disease course, pain intensity, orthodontic treatment history, previous medication history, tooth loss, and treatment method.

[0042] In step 2, the univariate analysis preliminarily screens relevant potential risk factors by comparing the result changes caused by any one of the 18 potential risk factors; using the risk factors preliminarily screened in the univariate analysis as independent variables and whether chronic TMD pain occurs as the dependent variable, multivariate Logistic regression analysis is performed to screen out the final risk factors.

[0043] The potential risk factors screened by the univariate analysis include: oral behavior, somatic symptoms, depression, anxiety, sleep quality, diagnostic classification, pain intensity; The final risk factors screened by the multivariate Logistic regression analysis include: somatic symptoms, anxiety, sleep quality, pain intensity.

[0044] The results of the multivariate Logistic regression analysis are shown in Table 1 below:

[0045] Note: OBC: Oral Behavior Checklist; PHQ-15: Patient Health Questionnaire-15; PHQ-9: Patient Health Questionnaire-9; GAD-7: Generalized Anxiety Disorder Questionnaire-7; PSQI: Pittsburgh Sleep Quality Index; 95% CI: 95% Confidence Interval; *: P < 0.05; **: P < 0.01 The second column, the B value, which is the coefficient of each independent variable at different classification levels in the model. The positive or negative sign indicates whether they have a positive or negative proportional relationship with the dependent variable. The third column, the P value, is the significance value of the Wald test. If it is less than 0.05, it indicates that the coefficient of the corresponding independent variable has statistical significance and has a significant impact on the change of different classification levels of the dependent variable. The fourth column, the OR value, also known as the odds ratio, mainly refers to the ratio P / (1 - P) in the case group divided by the ratio P / (1 - P) in the control group. It represents the degree of influence of one variable on another variable in the relationship between two categorical variables. Specifically, if the OR value is greater than 1, it means that the first variable is more likely to be associated with the occurrence of the event relative to the second variable; if the OR value is less than 1, it means that the first variable is more likely to be associated with the non-occurrence of the event relative to the second variable.

[0046] In step three, based on the Logistic regression equation, the formula for calculating the Odds of chronic pain occurring after non-invasive treatment is expressed as follows: Odds = EXP[A + 0.690×B + 1.472×C + 3.508×D + E + 1.094×F + 1.707×G + 2.812×H + 0.815×I + 0.775×J - 5.495]; Among them, A represents the assignment value when the PHQ-15 score is 0 - 4 points, B represents the assignment value when the PHQ-15 score is 5 - 9 points, C represents the assignment value when the PHQ-15 score is 10 - 14 points, D represents the assignment value when the PHQ-15 score ≥ 15 points, E represents the assignment value when the GAD-7 score is 0 - 4 points, F represents the assignment value when the GAD-7 score is 5 - 9 points, G represents the assignment value when the GAD-7 score is 10 - 14 points, H represents the assignment value when the GAD-7 score ≥ 15 points, I represents the assignment value of the PSQI score, and J represents the assignment value of the pain intensity.

[0047] The incidence rate of chronic pain P after non-invasive treatment = Odds / (1 + Odds) × 100%.

[0048] The assignment rules involved in the Odds calculation formula are as follows: The PHQ-15 score is assigned 0 when it is 0 - 4, 1 when it is 5 - 9, 2 when it is 10 - 14, and 3 when it is ≥15; the GAD-7 score is assigned 0 when it is 0 - 4, 1 when it is 5 - 9, 2 when it is 10 - 14, and 3 when it is ≥15; the PSQI score is assigned 0 when it is 0 - 5, and 1 when it is >5; the NRS score of the baseline pain intensity is assigned 0 when it is 0 - 3, and 1 when it is >3.

[0049] In a specific embodiment, if a patient's PHQ-15 score for somatization degree is 9, GAD-7 score for anxiety degree is 10, PSQI score for sleep quality is 6, and NRS score for baseline pain intensity is 3, then Odds = EXP(0.690 + 3.414 + 0.815 + 0 - 5.495) = 0.562. The probability of the patient having chronic pain after non-invasive treatment is 0.562 / (1 + 0.562)×100% = 36.0%.

[0050] To improve the visibility of the model, the present invention visualizes the prediction model in the form of a nomogram, as Figure 1 shown. According to the nomogram, the "scores" corresponding to each predictive factor can be obtained. The sum of these scores is the "total score", and the predictive probability corresponding to the total score is the risk level of an individual having chronic TMD pain.

[0051] Figure 1 In this, PHQ-15: Patient Health Questionnaire-15; GAD-7: Generalized Anxiety Disorder Scale; PSQI: Pittsburgh Sleep Quality Index.

[0052] The present invention evaluates the discrimination of the model through the C-index and the area under the receiver operating characteristic curve (ROC) (area under curve, AUC), the calibration curve of Calibration to evaluate the calibration of the model, and the Decision Curve Analysis (DCA) to evaluate the clinical practicability of the model.

[0053] (1) C-index: The C-index, i.e., the concordance index, represents the probability that the prediction result is consistent with the actual observation result. The value range is 0.5 - 1. 0.5 indicates completely random and the model has no discrimination, and 1 indicates complete consistency. Generally, it is considered that when 0.5 < C-index < 0.7, it indicates that the discrimination of the model is low; when 0.7 < C-index < 0.9, it indicates good discrimination; when C-index > 0.9, it indicates high discrimination.

[0054] (2) AUC: The AUC is the area under the ROC curve. The higher the AUC, the better the discrimination of the model. If 0.5 < AUC < 0.7, it indicates that the discrimination of the model is low; if 0.7 < AUC < 0.9, the discrimination of the model is good; if AUC > 0.9, the discrimination of the model is high.

[0055] (3) Calibration calibration curve: The Calibration calibration curve can reflect the degree of consistency between the actual occurrence probability and the predicted probability. In an ideal situation, the Calibration calibration curve is a curve with an intercept of 0 and a slope of 1. The higher the degree of fitting between the waveform of the calibration curve and the waveform in the ideal situation, the more accurate it indicates.

[0056] (4) DCA: By quantifying the net benefit at different threshold probabilities in the TMD cohort, the clinical utility of the risk prediction model for chronic pain TMD after non-invasive treatment is determined. The more the net benefit, the better the clinical utility of the model.

[0057] In terms of discrimination, the C-index of the risk prediction model in the present invention is 0.858; the AUC is 0.829, see Figure 2 , both indicators are between 0.7 and 0.9, indicating that the discrimination of the risk prediction model is good, and the ability to distinguish whether chronic pain occurs after non-invasive treatment for TMD patients is strong.

[0058] Figure 2 is the ROC curve graph. In the graph, the X-axis is the false positive rate, representing the probability of the actual negative class in the predicted positive class; the Y-axis is the true positive rate, representing the probability of the actual positive class in the predicted positive class. The diagonal dotted line represents random guessing, and the model has no discrimination; the red line represents the ROC curve. The more the red line deviates from the diagonal, the better the discrimination of the model. The AUC is the area under the ROC curve.

[0059] In terms of calibration degree of the chronic pain risk prediction model of the present invention, in the calibration graph ( Figure 3 ), it shows that the Calibration calibration curve fits well with the ideal situation, and the curve shape is close to 45°, indicating that the calibration degree of this risk prediction model is good, and the consistency between the probability of predicting chronic pain after non-invasive treatment for TMD patients and the actual probability is relatively high.

[0060] Figure 3 is the calibration graph of the prediction model. In the graph, the X-axis is the predicted probability of chronic TMD pain risk, and the Y-axis represents the actually diagnosed chronic TMD pain (proportion). The diagonal dotted line represents the ideal model, and the solid line represents the prediction model. The higher the degree of fitting between the solid line and the diagonal dotted line, the better the calibration degree of the model.

[0061] In terms of clinical utility, in the DCA curve of the prediction model ( Figure 4 ), the curve is above the two extreme lines within the range of the threshold probability of approximately 10% - 98%, indicating good clinical utility of the risk prediction model.

[0062] Figure 4 This is a clinical decision curve graph. In the graph, the X-axis represents the threshold probability, and the Y-axis represents the net benefit. The red curve represents the risk prediction model for chronic pain TMD after non-invasiveness. The other two lines represent two extreme situations respectively: the horizontal line indicates that it is assumed that no chronic pain occurs after non-invasive treatment for all patients; the diagonal line indicates that chronic pain occurs for all patients after non-invasive treatment. The decision curve shows that when the threshold probability is within the range of approximately 10% - 98%, using this risk prediction model is more beneficial than using the extreme lines.

[0063] The present invention uses the Bootstrap resampling method to internally validate the model. The Bootstrap resampling method performs repeated sampling in the original cohort to construct a Bootstrap resampled sample, and uses the resampled sample as the training set and the original cohort as the validation set to evaluate the performance of the model. This study repeats the above process 1000 times. The C-index is used as the evaluation index to repeatedly test the prediction ability of the model. If the test results of the two times are similar, and the C-index obtained after internal validation still indicates good discrimination of the model, it shows that the model has passed the internal validation. In the present invention, the C-index obtained after internal validation (repeated sampling 1000 times) is 0.843, and the C-indexes obtained from the two tests before and after are similar (0.858, 0.843), and both are between 0.7 and 0.9, indicating that the model has passed the internal validation.

[0064] The early risk prediction system for chronic pain temporomandibular joint disorders provided by the present invention is as Figure 5 shown, including an input module, a conversion module, and an output module; the risk prediction system is loaded and embedded in a mobile applet; Figure 5 In

[0065] the input module, there are included the self-assessment questionnaires of PHQ-15, GAD-7, PSQI, and the pain score, which are respectively used to evaluate the somatization, anxiety, sleep quality, and baseline pain intensity of the patient - 4 risk factors related to chronic TMD pain screened from 18 potential risk factors through univariate and multivariate Logistic regression analysis. After the patient fills them out, three self-assessment questionnaire scores and a pain intensity score can be obtained.

[0066] The conversion module is limited. According to the assignment method described in the present invention (assignment method: PHQ-15 score: 0-4 points = 0, 5-9 points = 1, 10-14 points = 2, ≥15 points = 3; GAD-7 score: 0-4 points = 0, 5-9 points = 1, 10-14 points = 2, ≥15 points = 3; PSQI score: 0-5 points = 0, >5 points = 1; baseline pain intensity: NRS score 0-3 = 0, >3 points = 1), the PHQ-15 score, GAD-7 score, PSQI score and pain score are converted into corresponding assigned scores.

[0067] Then, according to the assignment situation, based on the Logistic regression analysis calculation formula, the risk probability of the patient developing chronic pain after non-invasive treatment is calculated.

[0068] For example, if the patient's PHQ-15 score is 12 points, GAD-7 score is 4 points, PSQI score is 7 points, and pain score is 4 points, then the corresponding assigned scores are: PHQ-15 score: 2 points; GAD-7 score: 1 point; PSQI score: 1 point; pain intensity: 1 point. Odds of developing chronic pain after non-invasive treatment = EXP[1.472 × 2 + 1.094×1 + 0.815×1 + 0.775×1 - 5.495] = 1.142, incidence rate of chronic pain after non-invasive treatment = 1.142 / (1 + 1.142) × 100% = 53.3%.

[0069] The output module outputs the results of "high risk" and "low risk" of the patient developing chronic pain after non-invasive treatment according to the risk probability of the patient developing chronic pain obtained by the conversion module. If the risk probability ≥50%, it is set as "high risk"; if the risk probability <50%, it is set as "low risk". For the convenience of the patient's understanding, in the output model, a visual processing of the risk prediction model - a nomogram ( Figure 1 ) is presented at the same time. The doctor can, according to the nomogram, more intuitively explain to the patient the origin of the risk probability, arouse the patient's attention, and thus be more cooperative with the subsequent individualized treatment.

[0070] Example 2

[0071] Patients with newly diagnosed temporomandibular joint disorders fill in the PHQ-15 (Somatization), GAD-7 (Anxiety), and PSQI (Sleep Quality) questionnaires in the mini-program and assess their pain intensity. If the patient's PHQ-15 score is 10, GAD-7 score is 6, PSQI score is 6, and they are diagnosed with moderate / severe pain, the conversion module assigns the above original scores as 2 points, 1 point, 1 point, and 1 point respectively according to the assignment method. Based on the Logistic regression analysis calculation formula, the risk probability of developing chronic TMD pain is calculated to be 53.3%, which is higher than 50%. The output module outputs the result as "High Risk" and presents the nomogram risk prediction model.

[0072] Through the risk prediction model and prediction system in the present invention, it can help a large number of doctors and therapists screen high-risk populations prone to chronic temporomandibular joint disorder pain and carry out individualized interventions in a timely manner. It can provide theoretical guidance for the early screening, individualized, and precise diagnosis and treatment of chronic temporomandibular joint disorder pain, improve the overall treatment level of temporomandibular joint disorders, reduce the occurrence of chronic pain, reduce the number of patient visits, and thus save medical resources, which has important social and economic significance.

[0073] In the present invention, different from calculating the risk level through complex formulas in the past, the present invention is presented in the form of a mini-program, making the model have the advantages of being convenient, intuitive, and easy to promote. In 488 subjects of the present invention, through the screening of 18 potential risk factors, somatization, anxiety, sleep quality, and baseline pain intensity are determined as risk factors for chronic pain in patients with temporomandibular joint disorders. Using Logistic regression analysis, a nomogram risk prediction model is constructed.

[0074] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented in various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0075] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0076] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0078] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0079] The protection scope of the present invention is not limited to the above embodiments. Without departing from the spirit and scope of the inventive concept of the present invention, all changes and advantages that can be conceived by those skilled in the art are included in the present invention, and the appended claims are used as the protection scope.

Claims

1. A method for establishing an early risk prediction model for chronic painful temporomandibular joint disorder, characterized in that: The establishment method comprises the following steps: Step 1: Collect biological, psychological, and social information of the subjects to obtain 18 potential risk factors; Step 2: Screen the potential risk factors using univariate analysis and multivariate logistic regression; Step 3: Establish and verify the risk prediction model based on the screened risk factors.

2. The method of establishing as claimed in claim 1, characterized in that: In step 1, the potential risk factors include: (1) Age: divided into 18-30 years old, 31-40 years old, 41-50 years old, 51-60 years old, and 61-70 years old according to each age group; (2) Gender: divided into male and female; (3) Body mass index: The patient's weight status is divided into four levels: underweight, normal, overweight, and obese; (4) Nationality: divided into Han and others; (5) Place of residence: divided into urban and rural areas; (6) Marital status: divided into married and single; (7) Oral behavior: divided into 2 groups according to the oral behavior checklist score, including: low oral behavior frequency, oral behavior checklist score 0-24 points; high oral behavior frequency, oral behavior checklist score ≥25 points; (8) Somatization: The patients were divided into 4 groups according to the Patient Health Questionnaire-15 score, including: no somatization, with a Patient Health Questionnaire-15 score of 0-4 points; mild somatization, with a Patient Health Questionnaire-15 score of 5-9 points; moderate somatization, with a Patient Health Questionnaire-15 score of 10-14 points; severe somatization, with a Patient Health Questionnaire-15 score of ≥15 points; (9) Depression: patients were divided into 4 groups according to the Patient Health Questionnaire-9 score, including: no depression, Patient Health Questionnaire-9 score 0-4 points; mild depression, Patient Health Questionnaire-9 score 5-9 points; moderate depression, Patient Health Questionnaire-9 score 10-14 points; severe depression, Patient Health Questionnaire-9 score ≥15 points; (10) Anxiety: The patients were divided into four groups according to the GAV-7 score, including: no anxiety, GAV-7 score of 0-4; mild anxiety, GAV-7 score of 5-9; moderate anxiety, GAV-7 score of 10-14; severe anxiety, GAV-7 score ≥15; (11) Sleep quality: The patients were divided into two groups according to the Pittsburgh Sleep Quality Index score: good sleep quality, Pittsburgh Sleep Quality Index score of 0-5 points; poor sleep quality, Pittsburgh Sleep Quality Index score of >5 points; (12) Diagnostic classification: divided into 3 groups: joint diseases, painful diseases and mixed diseases; (13) Disease course: divided into <1 month, 1-3 months and >3 months; (14) Pain intensity: pain is graded numerically, including: pain numerical rating score of 0-3 points, indicating no / mild pain; pain numerical rating score >3 points, indicating moderate / severe pain; (15) History of orthodontic treatment: divided into yes and no; (16) Previous medication history: divided into yes and no; (17) Tooth loss: divided into yes and no; (18) Treatment methods: divided into six groups, including health education, drug therapy, physical factor therapy, drug therapy + physical factor therapy, physical factor + manual therapy, drug + physical factor + manual therapy.

3. The establishment method according to claim 1, characterized in that: In step 2, the univariate analysis preliminarily screens potential risk factors by comparing the changes in results caused by changes in any one of the 18 potential risk factors; Multivariate Logistic regression analysis was performed using the risk factors initially screened in the univariate analysis as independent variables and the occurrence of chronic temporomandibular joint disorder pain as the dependent variable to screen out the final risk factors.

4. The establishment method according to claim 3, characterized in that: The potential risk factors screened out by the univariate analysis include: oral behavior, somatization, depression, anxiety, sleep quality, diagnostic classification, and pain intensity; The mathematical model of the multi-factor Logistic regression analysis is estimated by the maximum likelihood method, which is expressed as: ， Where P represents the probability of chronic temporomandibular joint disorder pain given a combination of risk factors; x i represents different risk factors; β0 represents the baseline log odds when all risk factors are 0; β j represents the regression coefficients of different risk factors; The final risk factors screened out by the multivariate logistic regression analysis included somatization, anxiety, sleep quality, and pain intensity.

5. The establishment method according to claim 1, characterized in that: In step 3, based on the logistic regression equation, the calculation formula for the odds of chronic pain after non-invasive treatment is as follows: Odds=EXP[A+0.690×B+1.472×C+3.508×D+E+1.094×F+1.707×G+2.812×H+0.815×I+ 0.775×J-5.495]; Among them, A represents the assignment of Patient Health Questionnaire-15 scores of 0-4 points, B represents the assignment of Patient Health Questionnaire-15 scores of 5-9 points, C represents the assignment of Patient Health Questionnaire-15 scores of 10-14 points, D represents the assignment of Patient Health Questionnaire-15 scores ≥15 points, E represents the assignment of Generalized Anxiety Questionnaire-7 scores of 0-4 points, F represents the assignment of Generalized Anxiety Questionnaire-7 scores of 5-9 points, G represents the assignment of Generalized Anxiety Questionnaire-7 scores of 10-14 points, H represents the assignment of Generalized Anxiety Questionnaire-7 scores ≥15 points, I represents the assignment of Pittsburgh Sleep Quality Index scores, and J represents the assignment of pain intensity; The incidence of chronic pain after non-invasive treatment P = Odds / (1+Odds) × 100%.

6. The establishment method according to claim 5, characterized in that: The assignment rules involved in the Odds calculation formula are as follows: A value of 0 was assigned when the Patient Health Questionnaire-15 score was 0-4, a value of 1 was assigned when the Patient Health Questionnaire-15 score was 5-9, a value of 2 was assigned when the Patient Health Questionnaire-15 score was 10-14, and a value of 3 was assigned when the Patient Health Questionnaire-15 score was ≥15; a value of 0 was assigned when the Generalized Anxiety Questionnaire-7 score was 0-4, a value of 1 was assigned when the Generalized Anxiety Questionnaire-7 score was 5-9, a value of 2 was assigned when the Generalized Anxiety Questionnaire-7 score was 10-14, and a value of 3 was assigned when the Generalized Anxiety Questionnaire-7 score was ≥15; a value of 0 was assigned when the Pittsburgh Sleep Quality Index score was 0-5, and a value of 1 was assigned when the Pittsburgh Sleep Quality Index score was >5; a value of 0 was assigned when the pain numerical rating score of pain intensity was 0-3, and a value of 1 was assigned when the pain numerical rating score of pain intensity was >3.

7. The establishment method according to claim 1, characterized in that: The discrimination, calibration and clinical practicality of the risk prediction model were evaluated by C index, calibration curve and DCA decision curve; the model was internally validated by Bootstrap method; and the risk prediction model was visualized by nomogram method.

8. An early risk prediction model for chronic painful temporomandibular joint disorder obtained by the establishment method as described in any one of claims 1 to 7.

9. An early risk prediction system for chronic painful temporomandibular joint disorder, characterized in that: The risk prediction system comprises the risk prediction model as claimed in claim 8; the risk prediction system comprises an input module, a conversion module, and an output module; The input module is used to fill in the Patient Health Questionnaire-15, Pittsburgh Sleep Quality Index, Generalized Anxiety Questionnaire-7 self-assessment questionnaire, assess pain intensity, and calculate questionnaire scores; The conversion module is used to convert the calculated questionnaire scores in the risk prediction model according to the assignment method, and calculate the risk probability of an individual developing chronic temporomandibular joint disorder pain according to the Logistic regression analysis mathematical model; The output module is used to convert the calculated risk level into an output result of high risk or low risk for output.

10. A method for early risk prediction of chronic painful temporomandibular joint disorder, characterized in that: The method comprises the following steps: Step 1, inputting the Patient Health Questionnaire-15 score, Generalized Anxiety Questionnaire-7 score, Pittsburgh Sleep Quality Index score and pain score of the subject to be predicted into the risk prediction system; Step 2: According to the scoring method and the Logistic regression analysis calculation formula, the risk prediction system processes and transforms the original data to obtain the risk probability of the subject developing chronic pain; Step 3: Classify high and low risk predictions and output high and low risk results and nomogram risk prediction model results.

11. Application of the establishment method as described in any one of claims 1-7, the risk prediction model as described in claim 8, the risk prediction system as described in claim 9, or the risk prediction method as described in claim 10 in the risk prediction of pain in temporomandibular joint disorder.

Images