The Establishment And Evaluation Of Prediction Model For Acute High-risk Chest Pain

Objective To establish and evaluate the risk prediction model of high-risk chest pain in patients with acute chest pain,and to guide the pre-examination and tria ge of patients with acute chest pain by using it.Methods A retrospective survey of patients with acute chest pain attending the e mergency department of the First Hospital of the University of Science and Tech nology of China from July 2020 to March 2021 was conducted to collect genera l demographic data,emergency consultation data,and emergency vital signs data of patients meeting the inclusion criteria.High-risk chest pain was used as an o utcome indicator,and its independent influencing factors were screened,and a bi nary logistic regression equation was established to construct a scoring model.T he area under the ROC curve（AUC）was used to evaluate its discriminatory abi lity,and the Hosmer-Lemeshow goodness-of-fit test and calibration curve were us ed to assess its calibration ability.The Jorden index was calculated to find the b est cut-off value for predicting high-risk chest pain.Results A total of 2506 patients with chest pain were included,and comparison between groups showed significant differences in age,gender,mode of admission,time of presentation,time of onset,systolic blood pressure,diastolic blood press ure,temperature,pulse,respiration,Sp O2,MEWS,SI,cardiac chest pain,non-car diac chest pain,non-specific chest pain（chest tightness,palpitations,dizziness,sy ncope,weakness,elevated blood pressure）,past history（hypertension,diabetes,co ronary artery disease,chronic kidney disease,chronic lung disease,cerebrovascula r disease,percutaneous coronary intervention,arrhythmia）were significantly differ ent（P<0.05）.Logistic regression analysis was performed using the stepwise forw ard method,and 14 independent influencing factors were included,including age,sex,mode of admission,Sp O₂,MEWS,SI,cardiac chest pain,noncardiac chest pain,nonspecific chest pain with elevated blood pressure,and past history（hyp ertension,diabetes,coronary heart disease,chronic kidney disease,arrhythmia）.A fter recalibration and assignment,an acute high-risk chest pain scoring model wa s constructed with a total score of-8 to 21.The AUC of the model for predicti ng high-risk chest pain was 0.876（95%CI:0.862～0.890）,with good discriminati on.Hosmer-Lemeshow goodness-of-fit test:P>0.05,good calibration curve fit:R²=0.9859.The best cut-off value for predicting high-risk chest pain was 7.5,whic h was used to distinguish whether the chest pain was high-risk or not;when the score was<0.5.When the score is<0.5,such patients can be basically exclude d from high-risk chest pain,and when the score is>14.5,such patients can basi cally be judged as having high-risk chest pain.Therefore,a score of<0.5 is reg arded as very low risk,0.5≤score≤7.5 is regarded as low risk,7.5<score≤14.5can be regarded as high risk,and a score>14.5 is regarded as very high risk.Patients with acute chest pain were classified into four levels:very low risk,low risk,high risk and very high risk,which corresponded to the four levels of em ergency pre-screening triage,in order to guide the pre-screening triage of patients with acute chest pain.Conclusion Based on clinical features,physiological indicators and other objective indicators,the acute high-risk chest pain risk prediction model has superior differentiation and calibration.By taking the cut-off value as the basis for grading,the grading is accurate,and the output of results is fast and objective,which has certain guiding value for acute chest pain pre-examination and triage and chest pain management.