Dataset Information

Comparison of Machine Learning Methods With National Cardiovascular Data Registry Models for Prediction of Risk of Bleeding After Percutaneous Coronary Intervention.

ABSTRACT:

Importance

Better prediction of major bleeding after percutaneous coronary intervention (PCI) may improve clinical decisions aimed to reduce bleeding risk. Machine learning techniques, bolstered by better selection of variables, hold promise for enhancing prediction.

Objective

To determine whether machine learning techniques better predict post-PCI major bleeding compared with the existing National Cardiovascular Data Registry (NCDR) models.

Design, setting, and participants

This comparative effectiveness study used the NCDR CathPCI Registry data version 4.4 (July 1, 2009, to April 1, 2015), machine learning techniques were used (logistic regression with lasso regularization and gradient descent boosting [XGBoost, version 0.71.2]), and output was then compared with the existing simplified risk score and full NCDR models. The existing models were recreated, and then performance was evaluated through additional techniques and variables in a 5-fold cross-validation in analysis conducted from October 1, 2015, to October 27, 2017. The setting was retrospective modeling of a nationwide clinical registry of PCI. Participants were all patients undergoing PCI. Percutaneous coronary intervention procedures were excluded if they were not the index PCI of admission, if the hospital site had missing outcomes measures, or if the patient underwent subsequent coronary artery bypass grafting.

Exposures

Clinical variables available at admission and diagnostic coronary angiography data were used to determine the severity and complexity of presentation.

Main outcomes and measures

The main outcome was in-hospital major bleeding within 72 hours after PCI. Results were evaluated by comparing C statistics, calibration, and decision threshold-based metrics, including the F score (harmonic mean of positive predictive value and sensitivity) and the false discovery rate.

Results

The post-PCI major bleeding rate among 3 316 465 procedures (patients' median age, 65 years; interquartile range, 56-73 years; 68.1% male) was 4.5%. The existing full model achieved a mean C statistic of 0.78 (95% CI, 0.78-0.78). The use of XGBoost and full range of selected variables achieved a C statistic of 0.82 (95% CI, 0.82-0.82), with an F score of 0.31 (95% CI, 0.30-0.31). XGBoost correctly identified an additional 3.7% of cases identified as high risk who experienced a bleeding event and an overall improvement of 1.0% of cases identified as low risk who did not experience a bleeding event. The data-driven decision threshold helped improve the false discovery rate of the existing techniques. The existing simplified risk score model improved the false discovery rate from more than 90% to 78.7%. Modifying the model and the data decision threshold improved this rate from 78.7% to 73.4%.

Conclusions and relevance

Machine learning techniques improved the prediction of major bleeding after PCI. These techniques may help to better identify patients who would benefit most from strategies to reduce bleeding risk.

SUBMITTER: Mortazavi BJ

PROVIDER: S-EPMC6624806 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:ImportanceDiagnostic coronary angiography in asymptomatic patients may lead to inappropriate percutaneous coronary intervention (PCI) due to a diagnostic-therapeutic cascade. Understanding the association between patient selection for coronary angiography and PCI appropriateness may inform strategies to minimize inappropriate procedures.ObjectiveTo determine if hospitals that frequently perform coronary angiography in asymptomatic patients, a clinical scenario in which the benefit of angiography is less clear, are more likely to perform inappropriate PCI.Design, setting, and participantsMulticenter observational study of 544 hospitals participating in the CathPCI Registry between July 1, 2009, and September 30, 2013.Main outcomes and measuresHospital proportion of asymptomatic patients at diagnostic coronary angiography and hospital rate of inappropriate PCI as defined by 2012 appropriate use criteria for coronary revascularization.ResultsOf 1 225 562 patients who underwent elective coronary angiography, 308 083 (25.1%) were asymptomatic. The hospital proportion of angiography among asymptomatic patients ranged from 1.0% to 73.6% (median, 24.7%; interquartile range, 15.9%-35.9%). By hospital quartile of asymptomatic patients at angiography, hospitals with higher rates of asymptomatic patients at angiography had higher median rates of inappropriate PCI (14.8% vs 20.2% vs 24.0 vs 29.4% from lowest to highest quartile, P < .001 for trend). This outcome was attributable to more frequent use of inappropriate PCI in asymptomatic patients at hospitals with higher rates of angiography in asymptomatic patients (5.4% vs 9.9% vs 14.7% vs 21.6% from lowest to highest quartile, P < .001 for trend). Hospitals with higher rates of asymptomatic patients at angiography also had lower rates of appropriate PCI (38.7% vs 33.0% vs 32.3% vs 32.9% from lowest to highest quartile, P < .001 for trend).Conclusions and relevanceIn a national sample of hospitals, performance of coronary angiography in asymptomatic patients was associated with higher rates of inappropriate PCI and lower rates of appropriate PCI. Improving preprocedural risk stratification and thresholds for coronary angiography may be one strategy to improve the appropriateness of PCI.

Project description:BackgroundPrediction of thrombotic and bleeding risk is important to optimize antithrombotic therapy after percutaneous coronary intervention.Methods and resultsWe developed the prediction rules for thrombotic and bleeding events separately in Japanese patients. Derivation and validation cohorts consisted of 4778 patients from CREDO-Kyoto (Coronary Revascularization Demonstrating Outcome Study in Kyoto) registry cohort 2 and 4669 patients from RESET (Randomized Evaluation of Sirolimus-Eluting Versus Everolimus-Eluting Stent Trial) and NEXT (Nobori Biolimus-Eluting Versus Xience/Promus Everolimus-Eluting Stent Trial). Primary thrombotic and bleeding events were a composite of myocardial infarction, definite or probable stent thrombosis or ischemic stroke, and GUSTO (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries) moderate or severe bleeding. The prediction rule for thrombosis assigned 2 points for severe chronic kidney disease, atrial fibrillation, peripheral vascular disease, and anemia and 1 point for age ?75 years, heart failure, diabetes mellitus, and chronic total occlusion. The prediction rule for bleeding assigned 2 points for thrombocytopenia, severe chronic kidney disease, peripheral vascular disease, and heart failure and 1 point for prior myocardial infarction, malignancy, and atrial fibrillation. In derivation and validation cohorts, area under the curve was 0.68 and 0.64, respectively, for thrombosis and 0.66 and 0.66, respectively, for bleeding. In the validation cohort, a high thrombosis risk score (?4, n=682) was associated with higher 3-year incidence of thrombotic events than a score that was intermediate (2-3, n=1178) or low (0-1, n=2809) (7.6%, 3.7%, versus 2.4%, respectively; P<0.0001). A high bleeding risk score (?3, n=666) was associated with higher incidence of bleeding than scores that were intermediate (1-2, n=1802) or low (0, n=2201) (8.8%, 4.1%, versus 2.3%, respectively; P<0.0001). Among 682 patients at high thrombotic risk, only 39 (5.7%) had low bleeding risk, whereas 401 (58.8%) had high bleeding risk with very high incidence of bleeding (11.6%).ConclusionsCREDO-Kyoto thrombotic and bleeding risk scores demonstrated modest accuracy in stratifying thrombotic and bleeding risks; however, a large proportion of patients at high thrombotic risk also had high bleeding risk.