Evaluation And Improvement Of Screening Model And Clinical Decision-Making For Coronary Artery Disease

Coronary artery disease(CAD)is one of the most common cardiovascular diseases in China.According to the data released by the National Center for Cardiovascular Diseases,the mortality rate of CAD in urban and rural residents continued to rise.The mortality rate reached 110/100,000 in 2015,nearly three times more than in 2012.The continuous increase in morbidity and mortality has brought a heavy health and economic burden to social development.As a preventable and curable disease,CAD screening and clinical decision-making are important parts to improve the prognosis of patients and reduce the disease burden.Previous studies have carried out a lot of exploration in the aspects of CAD screening and clinical decision-making.In terms of CAD screening,researchers have proposed a variety of risk prediction models for CAD.In 1979,Professor Diamond and Forrester established the classic Diamond-Forrester model using age,gender,and angina symptoms.This model was simple in clinical application and was a milestone work.Subsequently,in order to further improve the predictive performance of the model,researchers continued to incorporate new variables to establish optimization models(such as the CAD Consortium Clinical Score,Duke Clinical Score,etc.).These optimized models greatly improved the performance and effectiveness.In terms of clinical decision-making,a series of evidence and guidelines were constantly updated to promote the standardization of diagnosis and treatment decisions.For complex CAD,it is recommended to use the heart team to make patient-centered diagnosis and treatment decisions.Although the previous work has achieved fruitful results,we still need to realize that there are still areas that need to be evaluated and improved in terms of CAD screening and clinical decision-making rationalization.In CAD screening,on the one hand,traditional clinical models were all developed from western cohorts.The populations used for training and validation were different in clinical characteristics,race,and region.The effectiveness of the model in general populations remains unclear.On the other hand,the existed models incorporated a large number of clinical variables,which were difficult to apply to large-scale early screening.Efficient and simple screening tools are still needed.In clinical decision-making,on the one hand,although guidelines and evidence were constantly updated,the appropriateness of coronary revascularization in China is still lack of evidence;on the other hand,although the heart team was recommended to improve clinical decision-making,previous studies have prompted the non-reproducibility,thus the best practice protocol for the heart team still needs to be developed.To solve these problems,we conducted a series of studies to evaluate and improve CAD screening models and clinical decision-making.The first part is the evaluation and improvement of CAD screening model.In this part,we systematically evaluated the predictive efficacy of existed CAD screening models and established a deep learning-based CAD prediction algorithm by analyzing facial photos to further optimize CAD screening.The second part is the evaluation and improvement of clinical decision-making for CAD.In this part,we evaluated the appropriateness of coronary revascularization decision-making in Chinese patients with stable CAD and explored the optimal heart team decision-making practice scheme for complex CAD.The main findings are as follows.Part ?.Evaluation and Improvement of Coronary Artery Disease Screening Model?.The Accuracy of Various Models in Predicting Coronary Artery Disease:A Systematic ReviewObjective:A variety of coronary heart disease risk prediction models have been developed for disease screening,but due to the heterogeneity of the modeling and verification populations,the actual application value of each model is not yet clear.This study aimed to systematically review various models for predicting coronary heart disease and clarify their predictive power.Methods:Pubmed,Embase and China National Knowledge Internet were searched comprehensively by computer.We included studies which were designed to develop and validate predictive models of CAD.The studies published from inception to September 30,2020 were searched.Two reviewers independently evaluated the studies according to the inclusion and exclusion criteria and extracted the baseline characteristics and metrics of model performance.Results:A total of 30 studies were identified,19 externally validated diagnostic predictive models for CAD.17 models had external validation group with area under curve(AUC)>0.7.The AUC for the external validation of the traditional models,including Diamond-Forrester model,updated Diamond-Forrester model,Duke Clinical Score,CAD consortium clinical score,ranged from 0.49 to 0.87.The AUC for the external validation in Chinese population was 0.57-0.78.Conclusion:Most models have modest discriminative ability.The predictive efficacy of traditional models varies greatly among different populations.Model based on Chinese populations was still warranted.?.Feasibility of Using Deep Learning to Detect Coronary Artery Disease Based on Facial PhotoAims:Facial features were associated with increased risk of coronary artery disease(CAD),which was a potential screening method.We aimed to develop and validate a deep learning algorithm for detecting CAD based on facial photos.Methods:In this multi-center cross-sectional study,we enrolled patients undergoing coronary angiography or computed tomography angiography at nine Chinese sites.Patients were divided into model development group and test group according to the time of enrollment.We used facial pictures of patients in the training group to establish a deep convolutional neural network algorithm for predicting CAD(at least one?50%stenosis),and tested the algorithm in the test group.Model evaluation measurement were area under the receiver operating characteristic curve(AUC),sensitivity and specificity.Results:Between July 2017 and March 2019,5,796 patients from eight sites were consecutively enrolled and randomly divided into training(90%,n=5,216)and validation(10%,n=580)groups for algorithm development.Between April 2019 and July 2019,1,013 patients from nine sites were enrolled in test group for algorithm test.The AUC in the test group was 0.730(95%confidence interval,0.699-0.761).Using an operating cut point with high sensitivity(0.80),the algorithm had sensitivity of 0.80 and specificity of 0.54 in the test group;.The AUC for the algorithm was higher than that for the Diamond-Forrester model(0.730 vs.0.623,p<0.001)and the CAD consortium clinical score(0.730 vs.0.652,p<0.001).Conclusion:Our results suggested that a deep learning algorithm based on facial photos can assist in CAD detection in this Chinese cohort.This technique may hold promise for pre-test CAD probability assessment in outpatient clinics or CAD screening in community.Further studies to develop a clinical available tool are warranted.Part ?.Evaluation and Improvement of Coronary Artery Disease Decision-makingI.Appropriateness of Coronary Revascularizaiton for Patients with Stable Coronary Artery DiseaseAim:In 2016,in order to standardize the selection of indications for revascularization surgery,the national center for cardiovascular diseases drafted Chinese appropriate use criteria for coronary revascularization.This study aimed to assess the appropriateness of coronary revascularization in patients with stable coronary artery disease(CAD).Methods:We conducted a multi-center,cross-sectional study involving stable CAD patients with coronary lesion stenosis?50%in 4 big cardiac centers.The appropriateness of coronary revascularization was judged by Chinese appropriate use criteria(AUC)for coronary revascularization.Results:From August 2016 to August 2017,5875 patients were consecutively enrolled.According to Chinese AUC,18.1%(1064/5875)of overall patients decision making were classified as inappropriate,43.6%(2560/5875)were classified as uncertain and 38.3%(2251/5875)were classified as appropriate.Among 376 paitients undergoing coronary artery bypass graft(CABG),3.5%were classified as inappropriate.Among 3452 paitients undergoing percutaneous coronary intervention(PCI),20.9%(723/3452)were classified as inappropriate.Among 2047 paitients undergoing medical therapy,16.0%(328/2047)were classified as inappropriate.Inappropriateness was more likely to be found in patients without angina(p<0.001).Appropriateness varied significantly among cardiologists(5.3%-25.0%,n=42).Conclusions:In this large,multi-center study,we found 18.1%of overall patients decision making were classified as inappropriate,with substantial variation among cardiologists.20.9%of patients undergoing PCI and 16.0%of patients undergoing medical therapy were found to be inappropriate.CABG was found to have high levels of appropriateness.?.Exploring Optimal Heart Team Protocol for Complex Coronary Artery DiseaseBackground:Although heart team was recommended by guideline for decision-making in patients with complex coronary artery disease(CAD),the decision-making stability was lack of evaluation and optimal protocol remained unknown.We aimed to assess inter-team agreement for revascularization decision-making and related influenced factors,so as to provide recommendations for optimal protocol.Methods:A sequential,explanatory mixed method study was conducted,including(1)a cross-sectional study retrospectively enrolling patients with complex CAD and four heart teams to assess the inter-team decision-making agreement and(2)a qualitative study that semi-structurally interviewed all heart team members to analyze the potential factors associated with decision-making discrepancy.Primary outcome was kappa value of inter-team decision-making agreement.Inductive thematic analysis was used to generate themes and subthemes attributing to decision-making discrepancy.Integrating qualitative and quantitative data,we explained how each subtheme affected decision-making agreement and provided corresponding recommendations based on these explanations.Finally,we provided a detailed heart team protocol by integrating our recommendations,published experience and guideline.Patient sample size was precalculated and interviewee sample size was identified by theoretical saturation.Results:A total of 101 patients with complex CAD were randomly enrolled from a consecutive angiography registry.Sixteen specialists were invited and randomly established four heart teams to make decisions for enrolled patients.Inter-team decision-making agreement was moderate(kappa 0.582).Decision-making may be influenced at three themes(specialist quality;team composition;meeting process)and ten subthemes(decision thought;understanding of disease and evidence;understanding of other discipline;personality;learning curve;personnel quality;number of team members;discipline selection;ratio of different disciplines and meeting form).Recommendations at five levels were provided,including(1)specialist selection,(2)specialist training,(3)team composition,(4)team training and(5)meeting process.A detailed implementation protocol to establish and deploy a qualified heart team was generated.Conclusions:Agreement between heart teams for revascularization decision-making in patients with complex CAD was moderate.Five recommendations to improve heart team modality were provided based on factors associated with decision-making discrepancy.A detailed heart team implementation protocol came into being.Randomized controlled trial was warranted to further confirm the protocol.