Current scores predicted clinically-diagnosed AF with AUC ~0.67-0.71 but showed lower accuracy for screen-detected AF (AUC ~0.57-0.60) in a Chinese screening cohort.
Do existing clinical scoring systems accurately predict incident asymptomatic subclinical new AF in older patients undergoing opportunistic screening?
Currently available clinical scores like C2HEST and CHA2DS2-VASc have inadequate discriminatory ability for predicting subclinical asymptomatic new AF in older patients undergoing opportunistic screening.
Absolute Event Rate: 0% vs 0%
Abstract Background C2HEST, mC2HEST and HATCH scores have been validated to predict incident atrial fibrillation (AF). We aimed to compare performance of these scoring systems in predicting asymptomatic subclinical new AF. Methods Consecutive patients aged ≥65 years attending medical outpatient clinics were prospectively enrolled for AF screening using handheld single-lead ECG (AliveCor) from 12/2014 to 12/2017 (NCT02409654). Three cohorts were formed, screen-detected AF, clinically-diagnosed AF and no AF. Performance of C2HEST, mC2HEST, HATCH, CHADS2 and CHA2DS2VASC scores in predicting screen-detected and clinically-diagnosed AF were compared by area under the curves (AUC). Results Of 11,972 subjects enrolled, 2,238 (18.7%) had known AF at enrollment. Yield of screen-detected AF on initial screening was 2.3% (n=223/9,734). AF was clinically-diagnosed during follow-up in 2.3% (n=216/9,440) and during subsequent screening in 71 initially screen-negative patients. Ability to predict clinically-diagnosed AF was similar between scores: C2HEST (AUC of 0.67(95%CI: 0.66-0.68)), mC2HEST (0.68(95%CI: 0.67-0.69)), HATCH (0.71(95%CI: 0.70-0.72)), CHADS2 (0.68(95%CI: 0.67-0.69)) and CHA2DS2VASC (0.67(95%CI: 0.66-0.68)). Performance in predicting screen-detected AF was lower compared to clinically-diagnosed AF C2HEST (AUC of 0.60(95%CI: 0.57-0.64)), mC2HEST (0.59 (95%CI: 0.56-0.63)), HATCH 0.58(95%CI: 0.55-0.62), CHADS2 0.58(95%CI: 0.55-0.61), and CHA2DS2VASC 0.57 (95%CI: 0.54-0.61), respectively. Predictors derived from cohort who participated in screening were internally validated by k-fold cross-validation (k=n=11,972, i.e., Leave-One-Out cross-validation). Significant predictors of asymptomatic new AF included coronary artery disease (Odds Ratio (OR): 1.42, 95%CI:1.20-1.68), hypertension (OR: 2.75, 95% CI: 2.33-3.26) , age≥70 years (OR: 2.61, 95%CI: 2.16-3.14) , previous stroke (OR: 2.13, 95%CI:1.79-2.53) and heart failure (OR: 2.27, 95%CI: 1.87-2.76), with AUC of 0.78 for derivation and 0.74 for cross-validation. Conclusion Current available scores derived from clinically-diagnosed AF demonstrated inadequate discriminatory ability while predicting subclinical asymptomatic new AF. Predicting system derived from cohorts undergoing opportunistic screening might help identify target population of screening with higher diagnostic yield and refine screening strategy, warranted further study.
Sun et al. (Sat,) reported a other. Current scores predicted clinically-diagnosed AF with AUC ~0.67-0.71 but showed lower accuracy for screen-detected AF (AUC ~0.57-0.60) in a Chinese screening cohort.