Correlation Study On Risk Of Dental Extraction In Elderly Patients With Cardiovascular Diseases

Objective:The aim of this study was to establish a prediction model of risk for dental extraction in elderly patients with cardiovascular diseases(CVDs),and then transform it into an electronic decision-making tool to identify the at-high risk patients,thus adopting prevention and intervention measures in advance to avoid cardiac complications.Methods:This study was a retrospective,observational,clinical study.A total of 833 elderly patients with cardiovascular diseases who fulfilled the inclusion criteria were enrolled into study between August 2017 and July 2018,including a training set which contained 603 elderly patients and an independent test set which contained 230 elderly patients.And data regarding clinical parameters,laboratory tests,clinical examinations before dental extraction,and 1-week follow-up were retrieved.In our study,the feature selection was determined by using logistic regression with penalized LASSO variable selection.First,we selected the variables with a p-value < 0.05 as predictors by using univariate analysis.Second,the importance of these predictors was identified by the LASSO binary logistic regression to acquire their coefficients.Third,we excluded the predictors with estimated zero coefficients,and the adjusted ORs(odds ratio)of the rest of 15 predictors were computed by using multivariate logistic regression.Predictors with a p-value < 0.05 were included in the final prediction model.After screening out predictive factors,a prediction model was constructed by the random forest algorithm using a 5-fold cross-validation method.At the optimal cut-off value,the following performance metrics were obtained on the basis of the confusion matrix: accuracy,sensitivity,specificity,PPV(positive predictive value),NPV(negative predictive value),LR+(positive likelihood ratios),LR-(negative likelihood ratios),and AUROC(area under the receiver operating characteristic curve).In addition to the receiver operating characteristic(ROC)curve as a traditional evaluation indicator,a precision-recall(PR)curve was also used as an important indicator of the prediction ability of the evaluation model in order to address the imbalance of the data.Comparison to the traditional statistical logistic regression(LR)model,the predictive power of the prediction model was evaluated by an independent test set.In the basis of the developed prediction model,we established initially an electronic application(app)of assessment for dental extraction.Results:The training set of 603 participants,including 282 men and 321 women,had an average participant age of 72.38±8.31 years.Among this population,129 patients had cardiac complications.Using a feature-selection method,11 predictors for risk of cardiac complications were screened out,including age(OR,1.076;95% confidence interval [CI],1.042–1.112;p<0.001),systolic blood pressure(OR,1.053;95% CI,1.033–1.074;p<0.001),heart rate(OR,1.042;95% CI,1.019–1.066;p<0.001),number of impacted teeth(OR,3.496;95% CI,1.184–6.736;p<0.001),current angina(2 weeks;OR,6.519;95% CI,3.021–14.067;p<0.001),grade 3 hypertension(OR,3.085;95% CI,1.223–7.782;p=0.017),history of cardiac-pacemaker insertion(OR,3.35;95% CI,1.36–8.25;p=0.009),rheumatic heart disease(OR,4.873;95% CI,1.348–17.617;p=0.016),history of oral antibiotics use prior to dental extraction(OR,2.126;95% CI,1.219–3.709;p=0.008),atrial fibrillation(OR,3.229;95% CI,1.188–8.775;p=0.022),and pulmonary arterial hypertension(OR,2.414;95% CI,1.079–5.401;p=0.032).Then,a prediction model was constructed based on the RF algorithm by using a 5-fold cross-validation method.The RF model produced an AUROC score of 0.93 in the training set and LR model produced an AUROC score of 0.88.At the optimal cut-off value of 18.5%,the RF model produced mean values of 82%,90%,80%,55%,97%,4.44 and 0.13 for accuracy,sensitivity,specificity,PPV,NPV,LR+,LR-,respectively;the corresponding values in the LR model were 80%,81%,80%,52%,94%,3.98 and 0.23,respectively.The AUROC scores of the LR model were 0.80 and the corresponding AUPRC(area under the precision-recall curve)scores were 0.35 in the independent test set,while the AUROC scores of the RF model were 0.83 and the corresponding AUPRC scores were 0.56 in the independent test set.Comparing to LR model,RF model had superior predictive performance.Finally,we chose RF model as the prediction model for risk of elderly patients with CVDs undergoing dental extraction.In the basis of the developed prediction model,we established initially an electronic app of assessment for dental extraction.Conclusion: The purpose of this study was to initially establish an electronic app of assessment to predict the occurrence of cardiac complications in elderly patients with cardiovascular diseases undergoing dental extraction.The findings of our study were expected to aid physicians and dentists in making more informed decisions in preoperative clinical assessment,which had a strong significance of clinical practicability.