CN113488123B

CN113488123B - Method for establishing diagnosis time-effect-based COVID-19 triage system, system and triage method

Info

Publication number: CN113488123B
Application number: CN202110432543.0A
Authority: CN
Inventors: 陈一君; 简文华; 梁振宇; 关伟杰; 梁文华; 李时悦; 张挪富; 郑劲平; 何建行; 钟南山
Original assignee: First Affiliated Hospital of Guangzhou Medical University
Current assignee: First Affiliated Hospital of Guangzhou Medical University
Priority date: 2021-04-21
Filing date: 2021-04-21
Publication date: 2023-07-18
Anticipated expiration: 2041-04-21
Also published as: CN113488123A; CN116959653A; CN116959654A

Abstract

The invention discloses a method for establishing a diagnosis system of a COVID-19 based on diagnosis aging, a computer system established by the method and a related diagnosis method of the COVID-19. The computer system includes: the data receiving module is used for receiving information including the age, chronic medical history, symptom onset date and first visit date of the consultant; the data processing module is used for comparing the received information with preset preferential treatment conditions; the output module is used for highlighting the consultant meeting the preset priority consultation condition and prompting the priority consultation; wherein the preset preferential treatment condition is selected from one or more of the following conditions: (a) has chronic disease and is older than 61 years, (b) has chronic disease and symptoms onset day to first visit day is older than 4 days, and (c) has no chronic disease but is older than 51 years.

Description

Method for establishing diagnosis time-effect-based COVID-19 triage system, system and triage method

Technical Field

The present invention relates to diagnosis of disease, and more particularly, to a method for establishing a diagnosis-age-based diagnosis-19 diagnosis system, a diagnosis-age-based diagnosis-19 diagnosis system established by the method, and a related diagnosis-age-based diagnosis-19 diagnosis method.

Background

The treatment of patients with new coronaries should avoid the transition from mild to severe and critical conditions to the greatest extent, and the mortality rate will be high once it is converted to severe and critical conditions. In addition, in epidemic prevention and control, early detection, early discovery and early treatment are fundamental preconditions for avoiding the transition of a new coronary patient from light to severe and critical diseases, which is the serious importance of epidemic prevention. From an imaging examination point of view, several studies have recorded that patients at a visit or examination (imaging examination within 6-14 days after onset of symptoms) are more likely to have exacerbating abnormal lung CT manifestations and lesions. The relationship between the prognosis of covd-19 and the diagnostic age (time-efficiency) and the importance of covd-19 diagnostic age in different populations has to be analyzed more deeply to date.

Increasing the diagnostic timeliness of the covd-19 means that the time from onset of symptoms to diagnosis is shortened, including: the duration from the onset of symptoms to the first visit, as dominated by the patient, and the duration from the first visit to the definitive visit, as dominated by the medical facility. On the one hand, we need the patient to identify the earliest symptoms and visit as soon as possible, but the time interval varies from person to person at present. On the other hand, hospitals should increase diagnosis efficiency, and the time required for diagnosis of COVID-19 in patients and the number of times of nucleic acid detection of reverse transcription polymerase chain reaction (RT-PCR) are not the same at present. Some patients need more than 2 nucleic acid tests to confirm diagnosis, and most patients can confirm diagnosis within 2 days, up to 7 days.

It is important to improve diagnostic timeliness, but it can also put pressure on the public health system.

Therefore, the diagnosis and treatment system based on the diagnosis time effect of the COVID-19 has practical significance in the areas with insufficient public health resources and fragile public health systems, so that suspicious visitors of the COVID-19 can be efficiently diagnosed, and the delay diagnosis and treatment of high-risk patients with bad prognosis can be avoided. In addition, there is a need for methods and systems for establishing an efficient triage system for different crowd samples to establish a triage system suitable for specific situations in each country or region.

Disclosure of Invention

The invention analyzes the distribution and cut-off value of the related time length of the diagnosis aging, and carries out survival analysis and comparison on different diagnosis aging groups so as to make up the neglected aspect in other researches and better notarize the effect of diagnosis aging. The invention builds a model based on diagnosis timeliness and clinical characteristics significant in analysis, and finds out a high-risk group needing higher diagnosis timeliness. The index selection of the data receiving module and the node selection of the data processing module of the system are calculated by the diagnosis system establishment method.

Based on the research, the application provides a diagnosis time-effect-based COVID-19 triage system, and a method and a system for establishing the triage system for brand-new crowd samples. Accordingly, in one aspect, the present invention also provides a method of establishing a diagnostic system based on a diagnostic age-based covd-19, the method comprising: (a) Collecting information including age, chronic medical history, symptom onset date, first visit date, diagnosis date, admission date, bad ending, and earliest occurrence date of bad ending of the patient diagnosed by the COVID-19; (b) Calculating the diagnosis time-efficiency related time length of each diagnostician, wherein the units are calculated in days, and the time length comprises symptom onset date to diagnosis date, symptom onset date to first diagnosis date and first diagnosis date to diagnosis date; calculating the duration from the date of admission to the earliest occurrence date of bad ending, and directly excluding the case if the duration is negative or more than 31 days; (c) Determining 95% quantiles of the diagnosis time-effect related duration, and removing abnormal cases of data with the quantiles of which the value is greater than 95%; (d) The diagnosis aging-related time length is taken as an independent variable, survival data from whether bad ending occurs or not and from the date of admission to the earliest occurrence date of bad ending is taken as a dependent variable, the diagnosis aging-related time length is determined to be converted into an optimal cut-off value (cutoff) of a two-class variable through Maxstat maximum statistics (maximally selected rank statistics) analysis, and the original diagnosis aging-related time length variable which is a continuous variable is directly replaced by the two-class variable, so that the influence of the extreme or abnormal aging-related time length on the stability of a diagnosis model is reduced, the variable effect is enhanced, and the application popularization of the diagnosis model is enhanced; (e) Taking diagnosis aging related time length, age, sex, smoking history and chronic disease history after the classification transformation as independent variables, taking survival data of whether bad ending happens and the time length from the date of admission to the earliest occurrence date of the bad ending as dependent variables, and carrying out single-variable log-rank test to obtain log-rank test p values of the independent variables respectively; (f) The method comprises the steps of taking variables, age, gender, smoking history and chronic disease history, including diagnosis aging related time length after classification conversion and logarithmic rank test p value which are lower than a set threshold value, as independent variables together, taking survival data of whether bad ending occurs or not and time length from the date of admission to the earliest occurrence date of the bad ending as dependent variables, and carrying out multivariate Cox proportion risk regression survival analysis to obtain Cox regression p values of the respective variables; (g) Taking variables including diagnosis time-effect related time length after classification conversion and Cox regression p value lower than a set threshold value as independent variables, and taking survival data of whether bad ending occurs or not and the time length from the date of admission to the earliest occurrence date of the bad ending as dependent variables, and carrying out conditional inference tree analysis to obtain a decision tree model; (h) And establishing a diagnosis system for the COVID-19 according to the obtained decision tree model.

In a specific embodiment, for the Maxstat maximum statistic method in step (d), the minprop (the minimal proportion of observations per group, i.e. the minimum ratio observed for each group) parameter value is set to 0.15. In a specific embodiment, for the Maxstat maximum statistic method in step (d), the break.time.by parameter value is 5.

In a specific embodiment, for the Maxstat maximum statistic method in the step (d), a surviving_point () function is used to calculate in R language programming, a minupp parameter value is set to 0.15, a prograsbar parameter value is set to TRUE, a variables parameter value is named as "timeston", a data parameter is used to call a data frame structure data set containing living data (time, event) and continuous variables to be segmented, a time parameter calls a living time index, is set to a time required for admission to poor end occurrence, an event parameter calls a living end index, is set to a state of whether poor end occurs, and an event value is set to 0 or 1.

In a specific embodiment, for the Maxstat maximum statistics method in the step (d), dividing each variable value according to the cut point (cut point) returned by the above survivin_cut point () function through the survivin_category () function, updating the variable value of "timedtod", and then obtaining the output cut point value, that is, the optimal cut value cutoff through the summary () function.

In one embodiment, the cutoff validity is verified for the Maxstat maximum statistic method in step (d). Constructing a survival function through a survivin () function, wherein the setting of time, event, data is the same as that of the previous step; the parameter of the (-predictors) is set as the 'timetod' by using the 'timetod' index generated in the previous step; using a ggsurvplot () function, drawing a survival curve for the above function, wherein the data setting is unchanged, the pval parameter value is set to TRUE, the risk.table parameter value is set to TRUE, the conf.int parameter value is set to TRUE, the surv.medium.line parameter value is set to "hv", and the linetype parameter value is set to "strata"; setting the break.time.by parameter value as 5 to achieve the best drawing effect; the parameters of the style theme or color panel such as ggtheme, palette are not limited; and if the two groups of survival curves do not have staggered trend, the values of the two classification cut points (cutoff) of the diagnosis aging related indexes are effective.

In one embodiment, for the univariate log rank test in step (e), a survival function is constructed by the survivin () function, the setting of time, event, data being the same as the previous step; each variable enters a parameter of the 'predictors' one by one; the above-mentioned surviving functions of surviving () are incorporated one by one using the surviving () function, carrying out logarithmic rank test one by one to obtain a p value; wherein the rho parameter value is set to 0 and the timefix parameter value is set to TRUE.

In one embodiment, for the multivariate Cox proportional-risk regression survival analysis in step (f), the surviving function of the multivariate is further constructed using the surviving () function, the setting of time, event, data being the same as the previous step; all independent variables enter the model; and (3) taking the coxph () function into the surviving function of the surviving vector, and carrying out multivariate Cox proportion risk regression survival analysis to obtain an output value Pr (> |z|) so as to obtain Cox regression p values of the respective variables.

In one embodiment, for the multivariate Cox proportional hazards regression survival analysis in step (f), the regular. Ok parameter value is set to TRUE, the model parameter value is set to FALSE, the x parameter value is set to FALSE, the y parameter value is set to TRUE, the method parameter value is set to ties, and ties are set to "Efron" in three values, "breslow", "exact", and using Efron approximation as a default value, because Efron is more accurate in handling the ending time of binding, and is computationally efficient.

In one embodiment, for the conditional inference tree analysis in step (g), a surviving function of multiple variables is further constructed using the surviving () function, the setting of time, event, data being the same as the previous step; placing the independent variables meeting the conditions into a model; the above surviving Surv function is incorporated using ctree () function, and the continuous, truncated, ordered, nominal variable and multiple response variables are recursively partitioned in a conditional inference framework to build a decision tree model.

In one embodiment, for the conditional inference tree analysis in step (g), the subset parameter value is set to NULL, the weights parameter value is set to NULL, the xtrafo parameter value is set to ptrafo, the ytrafo parameter value is set to ptrafo, and the score parameter value is set to NULL.

In some embodiments, wherein the set threshold for the log rank test p-value and the set threshold for the Cox regression p-value are 0.05 to 0.1. In some embodiments, wherein the abnormal case is a case of one or more of information of age missing, chronic medical history, date of onset of symptoms, date of first visit, date of confirmed diagnosis, date of admission, poor outcome, earliest occurrence of poor outcome. In some embodiments, wherein the adverse outcome comprises one or more of ICU, severe pneumonia, invasive ventilation, death. In some embodiments, wherein the severe pneumonia is assessed according to the national acute severe pneumonia clinical practice expert consensus.

The present invention evaluates the risk of adverse consequences in a patient with covd-19 by layering diagnostic timeliness. The analysis was performed on severe cases assessed as severe pneumonia or in intensive care units/invasive ventilation/death according to the national institute of clinical practice for severe pneumonia. The cutoff value for diagnostic timeliness (defined by the duration from symptom onset to diagnosis, including the duration from symptom onset to first visit and the duration from first visit to diagnosis) was studied by maximum selection of rank statistics. A comparison between survival analysis (log rank test, COX regression, conditional inference tree) and different age groups was performed.

The inventors found from the above analysis that the median of the time period from the onset of symptoms to the diagnosis, from the onset of symptoms to the first visit, and from the first visit to the diagnosis was 6 days, 3 days, and 2 days, respectively, and the cutoff values were 5 days, 4 days, and 3 days, respectively. After adjustment of age, sex, smoking status and complications status, age (risk ratio [ HR:1.03;95% CI: 1.01-1.04), complications (HR: 1.84;95% CI: 1.23-2.73), duration from onset of symptoms to diagnosis >5 days (HR: 1.69;95% CI: 1.10-2.60); age (hazard ratio [ HR:1.03;95% CI: 1.01-1.04), complications (HR: 1.77;95% CI: 1.18-2.68), duration from onset of symptoms to first visit >4 days (HR: 1.56;95% CI: 1.07-2.26), are independent predictors of prognosis of COVID-19. It follows that the longer the time from onset of symptoms to diagnosis, the worse the prognosis of covd-19, especially the delayed visit. The CTREE model shows that diagnosing age, complications are important nodes. The delay in visit is >4 days from the onset of symptoms to the first visit, and differential analysis shows that patients characterized by men, elderly, dry cough, coughing with sputum, shortness of breath, COPD are more common in the delay visit group.

In another aspect, the invention provides a computer system for diagnosis and diagnosis of a COVID-19 based on diagnosis and aging, comprising: the data receiving module is used for receiving information including the age, chronic medical history, symptom onset date and first visit date of the consultant; the data processing module is used for comparing the received information with preset preferential treatment conditions, wherein the preferential treatment comprises one or more of preferential medical advice, preferential sampling, preferential sample delivery and preferential detection; the output module is used for highlighting the consultant meeting the preset priority consultation condition and prompting the priority consultation; wherein the preset preferential treatment condition is selected from one or more of the following conditions: (a) has chronic disease and is older than 61 years, (b) has chronic disease and symptoms onset day to first visit day is older than 4 days, (c) has no chronic disease but is older than 51 years.

In some implementations, the output module is configured to set the priority according to the following rules: (1) Setting a doctor suffering from chronic disease and older than 61 years old as a first priority; (2) Setting a consultant with chronic disease and symptoms onset day to first visit day greater than 4 days as a second priority; (3) Setting a doctor without chronic disease but older than 55 years old as a third priority; and (4) setting a fourth priority for a doctor having no chronic disease but an age of more than 51 years old. In some embodiments, the output module is further configured to: when the patient age is chronically ill but older than 51 years, the prompt is that: it is recommended that the diagnosis of the covd-19 nucleic acid test be provided within no more than 5 days from the date of symptom onset. In some embodiments, the output module is further configured to: when the consultant has no chronic disease but is older than 55 years of age, prompt: it is recommended that the diagnosis result of the covd-19 nucleic acid test be provided within not more than 3 days from the first visit.

In some embodiments, the history of chronic disease includes one or more of COPD, diabetes, hypertension, coronary heart disease, cerebrovascular disease, hepatitis b, tumors, chronic kidney disease, and immunodeficiency. The invention selects the chronic diseases based on the discharge diagnosis in the 1590 patient electronic medical records, calculates the frequency and the duty ratio of the combined chronic disease characteristics, obtains 15 chronic complications which occur at high frequency and high duty ratio, and selects the chronic complications which are visible in bad ending cases, wherein the bad ending comprises one or more of ICU, severe pneumonia, invasive ventilation and death, and the severe pneumonia is evaluated according to the national acute severe pneumonia clinical practice expert consensus.

In some embodiments, the symptoms include one or more of dry cough, pharyngalgia, conjunctival congestion, nasal obstruction, headache, expectoration, tiredness, hemoptysis, shortness of breath, nausea, vomiting, diarrhea, joint muscle soreness, chills, fever. The invention selects the symptoms, which is based on the symptoms records obtained by collecting the complaints and retrospective information of the admission records in the 1590 patient electronic medical records, calculates the frequency and the duty ratio, and selects the symptoms which occur and are stably reproduced at the first 15 high frequencies and the high duty ratio, thereby determining the definition of the symptoms.

The invention discovers through analysis by a method for establishing a diagnosis system that the median of the time periods from symptom onset to diagnosis, from symptom onset to first visit and from first visit to diagnosis are 6 days, 3 days and 2 days respectively, and the cut-off values are 5 days, 4 days and 3 days respectively. The CTREE model in the method is established to show that diagnosis of aging, age and complications is an important node.

Still another aspect of the present invention provides a diagnostic method of a covd-19 based on diagnostic timeliness, comprising: (S1) receiving information including age, chronic medical history, date of symptom onset, and date of first visit of the visit; (S2) comparing the received information with preset priority visit conditions, the priority visit including one or more of priority order, priority sampling, priority sample delivery, and priority detection; (S3) highlighting the consultant meeting the preset priority consultation condition, prompting the priority consultation; wherein the preset preferential treatment condition is selected from one or more of the following conditions: (a) has chronic disease and is older than 61 years, (b) has chronic disease and symptoms onset day to first visit day is older than 4 days, (c) has no chronic disease but is older than 51 years.

In some embodiments, step S3 sets the priority according to the following rule: (1) Setting a doctor suffering from chronic disease and older than 61 years old as a first priority; (2) Setting a consultant with chronic disease and symptoms onset day to first visit day greater than 4 days as a second priority; (3) Setting a doctor without chronic disease but older than 55 years old as a third priority; and (4) setting a fourth priority for a doctor having no chronic disease but an age of more than 51 years old.

In some embodiments, the step (S3) further comprises prompting when the visitor is chronically ill but older than 51 years of age: it is recommended that the diagnosis of the covd-19 nucleic acid test be provided within no more than 5 days from the date of symptom onset. In some embodiments, the step (S3) further comprises prompting when the visitor is chronically ill but older than 55 years of age: it is recommended that the diagnosis result of the covd-19 nucleic acid test be provided within not more than 3 days from the first visit date.

In some embodiments, the history of chronic disease includes one or more of COPD, diabetes, hypertension, coronary heart disease, cerebrovascular disease, hepatitis b, tumors, chronic kidney disease, and immunodeficiency. The chronic diseases are selected by calculating the frequency and the duty ratio of the combined chronic disease characteristics based on the discharge diagnosis in the 1590 patient electronic medical records, obtaining 15 chronic complications which occur at high frequency and high duty ratio, and selecting the chronic complications which are visible in bad ending cases, wherein the bad ending comprises one or more of ICU, severe pneumonia, invasive ventilation and death, and the severe pneumonia is evaluated according to the national acute severe pneumonia clinical practice expert consensus.

The diagnosis method and the system are provided based on analysis of a large number of admission records, key node information in the diagnosis method and the system is extracted, diagnosis can be efficiently performed on a doctor based on different requirements of diagnosis confirmation time, delayed diagnosis is avoided, prognosis of a patient is promoted, medical resources are reasonably distributed, and the value of diagnosis time research is fully exerted. Moreover, by summarizing the logic and methods used in the above-described studies, the inventors have also determined a method of establishing an efficient triage system by which triage systems suitable for specific situations of each country can be rapidly established using data of each country.

Drawings

Fig. 1: frequency density map of the duration of diagnostic related timelines and results.

Fig. 2: log rank statistics are selected for maximum time duration cut-off points from symptom onset to diagnosis. a) Patients are divided into two groups, high (right part) and low (left part) according to the duration of the patient from symptom onset to diagnosis; the cut-off point (dashed line, showing the highest point) is defined by the maximum selected rank statistic; b) The time-dependent risk of reaching severity between patients with a period of "high" (lower curve) from onset of symptoms to diagnosis and a period of "low" (upper curve) from onset of symptoms to diagnosis; the transparent part represents the 95% confidence interval (95% ci). The maximum selection rank statistic allows evaluation of the cut-off points, which divide the observations into two groups by consecutive or ordered prediction variables. Calculation of the exact distribution of the maximum selected rank statistic is discussed and a new distribution lower bound is derived based on an extension of the algorithm to the exact distribution of the linear rank statistic.

Fig. 3: the log rank statistic is selected at maximum from symptom onset to first visit and from first visit to time length cut-off point of confirmed diagnosis. a) The patients are divided into two groups, high (right part) and low (left part) according to the duration of the patient from symptom onset to first visit; the cut-off point (where the two lines intersect and where the dashed line shows) is defined by the maximum selected rank statistic; b) A time-dependent risk of reaching a severity between the patient from the onset of symptoms to the first visit of the time "high" (lower curve) and the time "low" (upper curve) from the onset of symptoms to the first visit of the time; the transparent portion represents the 95% confidence interval (95% ci); c) The patients are divided into two groups, high (right part) and low (left part) according to the duration of the patient from the first visit to the definitive visit; the cut-off point (where the two lines intersect and where the dashed line shows) is defined by the maximum selected rank statistic; d) A time-dependent risk of reaching a severity between patients with a "high" duration from the first visit to the confirmed diagnosis (lower curve) and a "low" duration from the first visit to the confirmed diagnosis (upper curve); the transparent part represents the 95% confidence interval (95% ci). The maximum selection rank statistic allows evaluation of the cut-off points, which divide the observations into two groups by consecutive or ordered prediction variables. Calculation of the exact distribution of the maximum selected rank statistic is discussed and a new distribution lower bound is derived based on an extension of the algorithm to the exact distribution of the linear rank statistic.

Fig. 4: total survival curve over 31 days.

Fig. 5: a conditional inference tree model of the prognosis of covd-19 with diagnostic timeliness and pre-hospital factors. a) Dividing the time-dependent risk reaching the severity into 5 parts according to the nodes which are obviously separated in the model tree; calculating p-values by means of a corresponding time-series test (log-rank test); the model includes duration from symptom onset to diagnosis, age, and chronic complications condition; b) Dividing the time-dependent risk reaching the severity into 5 parts according to the nodes which are obviously separated in the model tree; calculating p-values by means of a corresponding time-series test (log-rank test); the model includes the duration from symptom onset to first visit, the duration from first visit to confirmed visit, age, chronic complications condition.

Detailed Description

Definition of the definition

Throughout the specification and claims, unless the context indicates otherwise, the word "comprise" and variations such as "comprises" and "comprising" will be understood to imply the inclusion of a stated integer, step or component but not the exclusion of any other integer, step or component. The term "comprising" when used herein may be replaced with the term "including" or "containing" or sometimes the term "having" is used herein.

As used herein, the term "covd-19" refers to novel coronavirus pneumonia (Corona Virus Disease 2019, covd-19), abbreviated as "novel coronavirus pneumonia", designated by the world health organization as "2019 coronavirus disease", refers to pneumonia caused by 2019 novel coronavirus infection.

As used herein, the term "triage" refers to the process of judging the severity and urgency of a patient's condition based on the patient's main symptoms and signs, medical history, etc., and its subordinate specialty, and simply and rapidly evaluating and classifying the type and severity of the condition, so that the patient gets proper diagnosis and care at the proper time and proper treatment area for the proper reasons according to triage priority. In the present invention, patients in need of priority visit can be determined by triage. Preferential treatment generally includes preferential prescribing, preferential sampling, preferential sample delivery, and preferential detection.

As used herein, the term "diagnostic timeliness" refers to the time efficiency of a patient at a critical point in time or event in the disease diagnosis process. Typically, these critical points in time or events include symptom onset, first visit, definitive diagnosis (i.e., obtaining a diagnosis), and the like. Accordingly, the term "diagnosis-age-related duration" may include a duration from the onset of symptoms to the diagnosis, a duration from the onset of symptoms to the first visit, and a duration from the first visit to the diagnosis. The duration is in particular in "days" in the present invention.

As used herein, the term "prognosis" refers to an empirically predicted disease progression. Disease prognosis is the understanding of a disease, and in addition to its clinical manifestation, assays and imaging, etiology, pathology, rules of illness, etc., it is important to evaluate the recent and distant efficacy, return to recovery, or extent of progression of a disease based on the timing and methodology of treatment in combination with new conditions found in the treatment regimen. The prognosis of a disease is related to many factors such as the timing of treatment of the patient, the extent of occurrence of the disease, the medical level, the combined disease, the personal ability of the doctor, the constitution, the age, whether the patient is looking at the disease or cognition for the disease, whether to continue the treatment, etc., and can vary greatly even if the same treatment is received. Some of the above factors are not resistant to unchangeable factors such as age, basal conditions, etc.; some can be improved by early intervention, such as early discovery, forward looking disease and early intervention and early treatment, which are all favorable for the prognosis to develop toward a good direction.

As used herein, the term "disease outcome" refers to the outcome of a disease that progresses to a degree or stage. In the present invention, "bad ending" may refer to the development of the COVID-19 to a severe degree. The severity may include a condition assessed as severe pneumonia or as having entered an intensive care unit/invasive ventilation/death according to the national institute of clinical practice for severe pneumonia.

As used herein, the term "Chinese emergency severe pneumonia clinical practice expert consensus" refers to the 2016 year of urgency by the Chinese society of physiciansThe medical doctors can issue' the national institute of clinical practice for severe pneumonia in emergency. In particular, according to the consensus, the definition criteria for cases of severe pneumonia (which should also be correspondingly shifted to ICU close observation and active treatment) are as follows: one of the primary criteria is met or more than (or equal to) 3 secondary criteria, wherein the primary criteria: (1) receiving invasive ventilation; (2) infective shock, the need for vasoactive drugs after fluid resuscitation; secondary standard: (1) respiratory rate increases (30 times/min); (2) partial pressure of arterial oxygen (PaO) ₂ ) Partial pressure of oxygen (FiO) ₂ ) Less than or equal to 250mmHg (1 mmHg=0.133 kPa); (3) pulmonary imaging shows multi-lobe infiltration; (4) a disturbance of consciousness and/or a disturbance of orientation; (5) the urea nitrogen in blood is more than or equal to 7mmol/L; (6) hypotension requires fluid resuscitation.

The present invention uses the term "maximum selection rank statistic" and "cut-off value". In the ROC curve, the dependent variable must be a binary variable, but sometimes we may face other situations that the dependent variable is survival data, quantitative data, etc., and the ROC curve is unable to function. In this case, it may be considered to find the cutoff value using the maximum selected rank statistic. Specifically, it is assumed that there are one dependent variable y (y may be classified material, continuous material, or living material) and one independent variable x. The maximum selection rank statistic corresponds to a separate division of each value of the x-variable, each division dividing the data into two groups, and simultaneously calculating a normalized statistic. The normalization statistic varies depending on the type of dependent variable, but in general, it reflects the difference between the two groups divided by a certain value. After all the divisions, a number of normalized statistics can be obtained, the largest of which is found, the corresponding division value being the best cut value (intrinsic or binary class variable), i.e. "cut-off value" as used herein.

In the technical scheme of the invention, the continuous variable of three time periods (the time period from symptom onset to diagnosis, the time period from symptom onset to first visit and the time period from first visit to diagnosis) can be changed into a classification variable through maximum selection rank statistics. Because of the small difference of one day and the large clutter, the conversion to a two-class variable results in a more stable analysis, and the cut-off value of the two classes of this continuous variable is referred to herein as the cut-off value. The best cut value is then found using the method described above. And then incorporated into the later model calculations.

Examples

Design, data source and data extraction

According to the provisional guidelines of the world health organization, these cases are confirmed by real-time RT-PCR analysis or high throughput sequencing of pharyngeal swabs and nasal swab specimens. Clinical data is checked and extracted as a computerized database by an experienced respiratory clinician and validated by double entry prior to analysis.

Data collection

Demographic, clinical, prognostic characteristics (i.e., gender, age, smoking status, primary past chronic disease, symptoms), primary exam at admission, and chest outcome were collected. For chronic complications, chronic Obstructive Pulmonary Disease (COPD), diabetes, hypertension, coronary heart disease, cerebrovascular disease, hepatitis b, malignant tumor, chronic kidney disease, immunodeficiency are recorded.

The date of symptom onset, date of first visit and date of diagnosis are also recorded, the time between the date of symptom onset and date of first visit and between the date of first visit and date of diagnosis is calculated, and if a number of less than 1.0 day appears, it is deleted from the analysis.

Patients who reached severity (as an endpoint indicator) were analyzed, including those assessed as severe pneumonia or admitted to an Intensive Care Unit (ICU)/invasive ventilation/death according to the national institute of clinical practice for severe pneumonia, the first day of reaching one of which determines the length of time to reach severity.

The prognosis and the length of time to the end of admission are recorded, but when the time length is negative, both will be deleted. Patients with a result-related duration exceeding 31 days will be excluded from the prognosis-related analysis along with patients with a result-deficiency.

Statistical analysis

The invention summarizes demographic information of all samples and compares data for severe and non-severe cases. Descriptive analysis of diagnosis timeliness (time period from symptom onset to diagnosis, time period from symptom onset to first visit, time period from first visit to diagnosis) and time to live (time period from admission to severity) was performed and frequency density plots were plotted. Finally, demographic and clinical data between high and low diagnostic timelines were also compared. The continuous data are expressed as mean and range. Non-parameter values were compared using a Wilcoxon rank sum test. Classification data are expressed in counts and percentages and compared using chi-square and Fisher exact tests.

To convert continuous variables into categorical variables, the disproportionate impact of extremes is reduced to stabilize the computational model ("days" as a measure, too accurate for assessing prognosis), a maximum selection rank statistic (Maxstat) analysis is performed to determine the appropriate cutoff point for diagnosing age. The Kaplan-Meier (KM) method was used to draw survival curves for different diagnostic time groups. In Maxstat analysis, a survivin_setpoint () function is used for calculation in R language programming, a minprop parameter value is set to 0.15, a prograsbar parameter value is set to TRUE, a variables parameter value is named as 'timetod', a data parameter is used for calling a data frame structure data set containing survival data (time, event) and continuous variables to be segmented, a time parameter is used for calling a survival time index, time required for admission to occurrence of bad results is set, an event parameter is used for calling a survival end index, a state of whether bad results occur is set, and an event value is set to 0 or 1. And dividing each variable value according to a cutting point (cut point) returned by the above-mentioned survivin_cut point () function through the survivin_category () function, updating the variable value of 'timetod', and then obtaining the output cut point value through the survivin_category () function, namely the optimal cut-off value cutoff. Validation of cutoff: constructing a survival function through a survivin () function, wherein the setting of time, event, data is the same as that of the previous step; the parameter of the (-predictors) is set as the 'timetod' by using the 'timetod' index generated in the previous step; using a ggsurvplot () function, drawing a survival curve for the above function, wherein the data setting is unchanged, the pval parameter value is set to TRUE, the risk.table parameter value is set to TRUE, the conf.int parameter value is set to TRUE, the surv.medium.line parameter value is set to "hv", and the linetype parameter value is set to "strata"; setting the break.time.by parameter value as 5 to achieve the best drawing effect; the parameters of the style theme or color panel such as ggtheme, palette are not limited; and if the two groups of survival curves do not have staggered trend, the values of the two classification cut points (cutoff) of the diagnosis aging related indexes are effective.

Total survival curves for patients with severity are plotted. The prognostic significance of all pre-admission factors (duration, sex, age, smoking status, whether there was a past chronic disease) was analyzed by log rank test and multivariate Cox proportional risk regression without violating the proportional risk hypothesis. HR and 95% confidence intervals (95% ci) are described. Wherein, for univariate logarithmic rank test, constructing survival function by survivin () function, setting time, event, data is the same as the previous step; each variable enters a parameter of the 'predictors' one by one; the above-mentioned surviving functions of surviving () are incorporated one by one using the surviving () function, carrying out logarithmic rank test one by one to obtain a p value; wherein the rho parameter value is set to 0 and the timefix parameter value is set to TRUE. Wherein, for multivariate Cox proportional risk regression survival analysis, a surviving function of the multivariate is further constructed by using a surviving () function, and the setting of time, event, data is the same as the previous step; all independent variables enter the model; and (3) taking the coxph () function into the surviving function of the surviving vector, and carrying out multivariate Cox proportion risk regression survival analysis to obtain an output value Pr (> |z|) so as to obtain Cox regression p values of the respective variables. Wherein, the sine. Ok parameter value is set as TRUE, the model parameter value is set as FALSE, the x parameter value is set as FALSE, the y parameter value is set as TRUE, the method parameter value is set as ties, and ties are set as "Efron" in three values of "Efron", "breslow", "exact", and using Efreon approximation as a default value, because Efreon is more accurate in processing the bound ending time, the calculation efficiency is higher.

Conditional inference tree analysis (Conditional inference tree analysis, CTREE) is a machine learning method performed as a complementary analysis with significance factors (probability value <0.1 for multivariate Cox proportional risk regression), which typically uses multiple significance tests and information measures (e.g., kunit) of feature permutations on tree nodes to partition the predictors most relevant to the results, and recursively split the tree, partitioning the patient into subsamples with different severity risk. CTREE recursively performs single variable splitting of dependent variables based on values over a set of covariates. CTREE tends to select variables with many possible splits or many missing values, using a saliency test program to select variables, rather than selecting variables that maximize an information metric (e.g., a coefficient of kunity). Wherein, using the survivin () function to further construct a multi-variable survival function, the setting of time, event, data is the same as the previous step; placing the independent variables meeting the conditions into a model; the above surviving Surv function is incorporated using ctree () function, and the continuous, truncated, ordered, nominal variable and multiple response variables are recursively partitioned in a conditional inference framework to build a decision tree model. Wherein the subset parameter value is set to NULL, the weights parameter value is set to NULL, the xtrafo parameter value is set to ptrafo, the ytrafo parameter value is set to ptrafo, and the score parameter value is set to NULL.

Meanwhile, the invention compares the prognosis characteristics between patients who visit within 4 days after symptom onset and other patients with the initial check result of admission. Probability values (P-value) <0.05 were considered statistically significant and were statistically analyzed using R software (R version 4.0.0 https:// www.r-project. Org /).

Results

Demographic and clinical characteristics

A summary of demographic information about the entire cohort is shown in table 1 and compares data between severity and non-severity.

Table 1 demographic and clinical characteristics of patients stratified by different severity levels

/>

Data are mean ± standard deviation, or median of band range, n (%), where n is the number of patient samples and% is the proportion of available data. COPD = chronic obstructive pulmonary disease.

Of these 1590, the median age was 48.0 years and only 647 patients (40.6%) were females. In addition to fever during or after hospitalization (88.0%), the most common symptom is dry cough (70.2%). 399 patients (25.1%) had at least one chronic established disease. More than 85% of patients have at least one chest CT or X-ray abnormality. Significant differences were also observed in diagnostic timeliness and result-related duration.

Diagnosing characteristics and cut-off points of time-lapse or result-related durations

The study further analyzed and described the diagnostic timeliness and the duration of the result correlation (table 2) and plotted the frequency density map (fig. 1). In these 1590 cases, the median time period from onset of symptoms to diagnosis, from onset of symptoms to first visit, from first visit to diagnosis, from admission to reach severity was 6, 3, 2, 8 days, respectively.

TABLE 2 diagnosis of age and duration of results

* In this queue, a total of 1590 cases, 1501 of which had symptom onset dates, 1512 had first visit dates, 935 had definitive diagnosis dates, 1246 had admission dates, 234 had progressed to severity and had definitive dates.

* According to the logit statistical method, the above 4 durations were completely random Missing (MCAR) considering age, past disease and endpoint factors.

To determine the cut-off value of the duration, 350 observations were deleted due to lack of duration of outcome (lack of date of admission or date of transition to severity of illness, or calculation of negative). In addition, 11 cases (5%, 11/237) were significantly outliers (much greater than the third quartile plus 1.5 times the difference between the first and third quartiles, authenticity could not be traced) and were excluded from the admission to reach severity. Thus, in the Maxstat analysis, a maximum of 1229 cases are retained to determine the optimal threshold.

The cut-off values for the time periods between symptom onset and diagnosis, between symptom onset and first visit, and between first visit and diagnosis were 5, 4, and 3 days, respectively, according to the optimal log rank test P-value (fig. 2, 3) (table 3).

TABLE 3 cut-off value and log rank test results for diagnostic timeliness

*<0.0005,**<0.0001

Prognosis analysis

Further prognostic analysis was performed from a total of 1229 cases with complete prognostic outcome indicators (date of admission, date of end, end result). Total survival curves were plotted (fig. 4). Univariate log rank analysis showed that age, gender, complications, duration associated with diagnostic aging were suspicious factors affecting the prognosis of covd-19 (P <0.05, table 4). After adjustment of age, sex, smoking status and complications status, multivariate Cox proportional risk regression analysis showed that age (risk ratio [ HR ]:1.03;95% ci: 1.01-1.04), complications status (HR: 1.84;95% ci: 1.23-2.73), duration from onset of symptoms to diagnosis (HR: 1.69;95% ci: 1.10-2.60) are strong independent predictors of the severity of covd-19. Age (HR: 1.03;95% CI: 1.01-1.04), complications status (HR: 1.77;95% CI: 1.18-2.68), duration from onset of symptoms to first visit (HR: 1.56;95% CI: 1.07-2.26) were significantly correlated with the severity of COVID-19 (Table 3) when considering the duration from onset of symptoms to first visit and from first visit to confirmed.

TABLE 4 univariate log rank analysis and multivariate Cox proportional hazards regression analysis

***＝p<0.001,**＝p<0.01,*＝p<0.05,#＝p<0.1

The relationship between prognosis results and significant factors (P < 0.1) in the multivariate COX regression was further analyzed by a conditional inference tree, determining the relationship between risk threshold and distinguishing overall survival. The CTREE model, taking into account age, chronic complications condition and duration from onset of symptoms to diagnosis, shows that past chronically ill patients aged over 61 years are more likely to reach severity. For patients without chronic complications, they may also be at similar risk to severity when diagnosed >5 days after onset of symptoms and aged over 51 years (fig. 5 a).

Another CTREE model, taking into account age, chronic comorbidities status, duration from onset of symptoms to first visit, and duration from first visit to confirmed, shows that no chronic comorbidities, confirmed >3 days after first visit, and the worst prognosis for patients over 55 years of age. Chronic comorbidities patients who were diagnosed >4 days after the onset of symptoms were ranked second (fig. 5 b). Both CTREE prognosis models show that diagnostically relevant age-related duration plays an important role in disease progression, especially in older and more complicated individuals.

In addition, by comparing the prognosis characteristics and admission primary test results of patients who have been diagnosed within 4 days after the onset of symptoms with those of other patients, it was found that the incidence of ARDS is high in patients who have been diagnosed for more than 4 days. The same group had a large number of patients (P < 0.05) characterized by men, elderly, dry cough, expectoration, shortness of breath, COPD (table 5).

TABLE 5 prognosis characteristics of patients with diagnosis and other patients within 4 days after onset of symptoms and initial hospitalization results comparison

Data are mean ± standard deviation, or median of band range, n (%), where n is the number of patient samples and% is the proportion of available data. ARDS = acute respiratory distress syndrome. DIC = disseminated intravascular coagulation. COPD = chronic obstructive pulmonary disease.

The invention is a study for deeply discussing the influence of diagnosis timeliness of the patient with the COVID-19 in China on prognosis and clinical characteristics for the first time. In the integrated risk assessment and prognosis survival model, pre-hospital factors such as complications, age, sex, smoking status, etc. of the covd-19 patient are also considered, along with the time period associated with diagnosis of age. The invention shows that diagnosis timeliness plays a critical role in the prognosis of the COVID-19, especially for elderly patients and chronic complications, and for patients with chronic obstructive pulmonary disease, the first visit is more likely to be delayed. According to the invention it is advantageous to reduce the total time between symptom onset and diagnosis to less than 5 days. Chronic complications patients suggest a visit within 4 days after the onset of symptoms. Patients over 55 years old should be diagnosed within 3 days after the visit. Thus, elderly patients and chronic complications patients are more susceptible to diagnostic timeliness. Notably, COPD patients are significantly more in the group with longer onset of symptoms to first visit, and COPD patients often present with symptoms of dry cough, expectoration, shortness of breath, etc. The present invention also finds more severe, worse prognosis or ARDS cases in the delayed visit group. In the case of covd-19 diagnosed by laboratory nationally, patients with long duration from onset of symptoms to diagnosis have a poor prognosis, especially those with delayed visits. Age-diagnosis shows that elderly patients are more sensitive and delayed diagnosis (e.g., more than 5 days after onset of symptoms) may deteriorate their prognosis. In addition, for elderly patients over 55 years old, a faster diagnosis should be made within 3 days. Patients with chronic complications should be diagnosed within 4 days after onset of symptoms. Proper typing should determine the risk stratification of patients by more carefully asking for medical history, identifying patients with delayed disease and patients more likely to develop poor prognosis, especially those with complications in the elderly. Patients suffering from COPD or any other respiratory disease should be more clear of their symptoms, seek medical attention in time, and be better guided.

Claims

1. A method of establishing a diagnostic-time-dependent spread-19 triage-based system, the method comprising:

(a) Collecting information including age, chronic medical history, symptom onset date, first visit date, diagnosis date, admission date, bad ending, and earliest occurrence date of bad ending of the patient diagnosed by the COVID-19;

(b) Calculating the diagnosis time-efficiency related time length of each diagnostician, wherein the units are calculated in days, and the time length comprises symptom onset date to diagnosis date, symptom onset date to first diagnosis date and first diagnosis date to diagnosis date; calculating the duration from the date of admission to the earliest occurrence date of bad ending, and directly excluding the case if the duration is negative or more than 31 days;

(c) Determining 95% quantiles of the diagnosis time-effect related duration, and removing abnormal cases of data with the quantiles of which the value is greater than 95%;

(d) The diagnosis aging related time length is taken as an independent variable, survival data from whether bad ending occurs or not and from the date of admission to the earliest occurrence date of the bad ending is taken as a dependent variable, the diagnosis aging related time length is determined to be converted into an optimal cut-off value of two classified variables through analysis of a Maxstat maximum statistic method, and the diagnosis aging related time length variable which is originally a continuous variable is directly replaced by the two classified variables;

(e) Taking diagnosis aging related time length, age, sex, smoking history and chronic disease history after the classification conversion as independent variables, taking survival data of whether bad ending occurs or not and the time length from the date of admission to the earliest occurrence date of the bad ending as dependent variables, and carrying out single-variable logarithmic rank test to obtain logarithmic rank test p values of the independent variables respectively;

(f) The method comprises the steps of taking variables, age, gender, smoking history and chronic disease history, including diagnosis aging related time length after classification conversion and logarithmic rank test p value which are lower than a set threshold value, as independent variables together, taking survival data of whether bad ending occurs or not and time length from the date of admission to the earliest occurrence date of the bad ending as dependent variables, and carrying out multivariate Cox proportion risk regression survival analysis to obtain Cox regression p values of the respective variables;

(g) Taking variables including diagnosis time-effect related time length after classification conversion and Cox regression p value lower than a set threshold value as independent variables, taking survival data of whether bad ending occurs or not and time length from the date of admission to the earliest occurrence date of the bad ending as dependent variables, and carrying out conditional inference tree analysis to obtain a decision tree model; and

(h) And establishing a diagnosis system for the COVID-19 according to the obtained decision tree model.

2. The method of claim 1, wherein the adverse outcome comprises one or more of ICU, severe pneumonia, invasive ventilation, death.

3. The method of claim 1, wherein the set threshold for the log rank test p-value and the set threshold for the Cox regression p-value are both 0.05 to 0.1.

4. The method of claim 1, wherein the abnormal case is a case of one or more of information of age of absence, history of chronic disease, date of onset of symptoms, date of first visit, date of diagnosis, date of admission, poor outcome, earliest occurrence of poor outcome.

5. The method of claim 2, wherein the severe pneumonia is assessed according to the national emergency severe pneumonia clinical practice expert consensus.

6. A computer system for diagnosis and diagnosis based on diagnosis and aging of covd-19, comprising:

the data receiving module is used for receiving information including the age, chronic medical history, symptom onset date and first visit date of the consultant;

the data processing module is used for comparing the received information with preset preferential treatment conditions, wherein the preferential treatment comprises one or more of preferential medical advice, preferential sampling, preferential sample delivery and preferential detection; and

The output module is used for highlighting the doctor who meets the preset priority doctor-seeing condition and prompting the priority doctor-seeing;

wherein the preset preferential treatment conditions are determined according to the method of any one of claims 1 to 5 and are selected from one or more of the following conditions:

(a) Has a chronic disease and is older than 61 years of age,

(b) Has chronic disease and symptoms are taken from the first treatment day to the first treatment day of more than 4 days, and

(c) No chronic disease but older than 51 years of age.

7. The triage computer system of claim 6, wherein the output module is further configured to set priority according to the following rules:

(1) Setting a doctor suffering from chronic disease and older than 61 years old as a first priority;

(2) Setting a consultant with chronic disease and symptoms onset day to first visit day greater than 4 days as a second priority;

(3) Setting a doctor without chronic disease but older than 55 years old as a third priority; and is also provided with

(4) The fourth priority is given to the patients who have no chronic disease but are older than 51 years old.

8. The triage computer system of claim 6, wherein the output module is further configured to: when the consultant is chronically ill but older than 51 years, a prompt suggests that the diagnosis of the covd-19 nucleic acid test be provided within no more than 5 days from the date of symptom onset.

9. The triage computer system of claim 6, wherein the output module is further configured to: when the consultant has no chronic disease but is older than 55 years, the prompt suggests that the diagnosis of the covd-19 nucleic acid test be provided within no more than 3 days from the first visit.

10. The triage computer system of claim 6, wherein the history of chronic disease comprises one or more of COPD, diabetes, hypertension, coronary heart disease, cerebrovascular disease, hepatitis b, tumor, chronic kidney disease, and immunodeficiency.

11. The triage computer system of claim 6, wherein the symptoms comprise one or more of dry cough, pharyngalgia, conjunctival congestion, nasal obstruction, headache, expectoration, tiredness, hemoptysis, shortness of breath, nausea/vomiting, diarrhea, joint muscle soreness, chills, fever.

12. A diagnostic method of covd-19 based on diagnostic timeliness, comprising:

(S1) receiving information including age, chronic medical history, date of symptom onset, and date of first visit of the visit;

(S2) comparing the received information with preset priority visit conditions, the priority visit including one or more of priority order, priority sampling, priority sample delivery, and priority detection;

(S3) highlighting the consultant meeting the preset priority consultation condition, prompting the priority consultation;

(a) Has a chronic disease and is older than 61 years of age,

(c) No chronic disease but older than 51 years of age.

13. The triage method according to claim 12, wherein step S3 sets priority according to the following rule:

14. The triage method according to claim 12, wherein in step S3, when the consultant has no chronic disease but is older than 51 years old, prompting: it is recommended that the diagnosis of the covd-19 nucleic acid test be provided within no more than 5 days from the date of symptom onset.

15. The triage method according to claim 12, wherein in step S3, when the consultant has no chronic disease but is older than 55 years old, prompting: it is recommended that the diagnosis result of the covd-19 nucleic acid test be provided within not more than 3 days from the first visit date.

16. The method of triage according to claim 12, wherein the history of chronic disease comprises one or more of COPD, diabetes, hypertension, coronary heart disease, cerebrovascular disease, hepatitis b, tumor, chronic kidney disease, and immunodeficiency.

17. The method of triage according to claim 12, wherein the symptoms include one or more of dry cough, sore throat, conjunctival congestion, nasal obstruction, headache, expectoration, tiredness, hemoptysis, shortness of breath, nausea/vomiting, diarrhea, joint muscle soreness, chills, fever.