18 Clinical risk model to predict likelihood of having good LV function post myocardial infarction. (May 2019)

• Record Type: Journal Article
• Authors: Wilson, David G; Routledge, Helen; Johnson, Tom; Harris, Jessica
• Is Part Of: Heart. Volume 105(2019)Supplement 6
• Page Start: A16
• Page End: A17
• Publication Date: 2019-05
• Subjects: Clinical risk score -- Echocardiography -- Acute myocardial infarction; Heart -- Diseases -- Treatment -- Periodicals; Cardiology -- Periodicals; 616.12
• DOI: 10.1136/heartjnl-2019-BCS.17
• Languages: English
• ISSNs: 1355-6037

Abstract

Abstract : Background: A national shortage of sonographers and NHS hospital beds challenges us in providing timely inpatient echocardiography to patients admitted to hospital with acute myocardial infarction (AMI). A clinical risk model to identify clinical predictors of having good LV function post AMI could potentially help risk-stratify patients for early discharge and expedited outpatient echocardiography. Aim: To develop a clinical risk score to predict the likelihood of having good LV function on transthoracic echocardiography post AMI. Methods: Data that had been collected for patients presenting to Worcestershire Acute Hospitals NHS Trust with AMI between July 2014 and November 2017 were used. These data had been collected as part of the Myocardial Infarction National Audit Project (MINAP). A clinical risk model was developed. Multiple imputation methods were used to deal with missing data. Logistic regression was used to determine to effect of these factors upon the outcome of good LV function (ejection fraction (LVEF). The results were externally validated with a MINAP dataset from the Bristol Heart Institute collected over the same time period. Results: A development dataset comprising 2232 patients was used; 931/1668 (42%) had good LV function. Factors entered in the model were decided a priori and were: site of infarction, previous myocardial infarction, heart rate, systolic BP, Killip class, peak troponin, ECG determining treatment, age and gender. The final model had an area under receiver operator curve c-statistic = 0.79 (95% CI 0.75, 0.82). The model was externally validated on 2238 patients from the Bristol Heart Institute. The discrimination of the model was moderate with c-statistic = 0.66 (95% CI 0.64, 0.68); 73% sensitivity and 59% specificity. Conclusion: This clinical risk score predicts the likelihood of having good LV function on transthoracic echocardiogram post myocardial infarction moderately well. Further work to improve the accuracy of the model could enable a move to earlier discharge and out patient echocardiography in more than a third of all AMI patients. Conflict of Interest: Nil