Development of a Multivariable Prediction Model to Estimate Probability of Sleep-Disordered-Breathing...

Sleep-disordered breathing (SDB) is common in patients with atrial fibrillation (AF) and negatively impacts treatment outcomes. Optimal tools for AF patient selection for SDB testing are lacking.

This study sought to develop and validate a prediction tool to detect patients who have AF with moderate-to-severe SDB.

Prospectively collected data on 442 consecutive ambulatory patients with AF who were undergoing polysomnography were used as the derivation sample. Performance was externally validated on a test cohort of 409 patients. Significant SDB was defined as an apnea-hypopnea-index ≥15/h. Multivariable logistic regression was used to construct a prediction model and calculate individual SDB probabilities.

Significant SDB was present in 34% and 54% of patients in the derivation and validation cohorts, respectively. The prediction model comprised age, sex, body mass index (BMI), diabetes, and previous stroke or transient ischemic attack. Following calibration, the model had a good discrimination ability for significant SDB on external validation (C-statistic: 0.75; 95% CI: 0.71-0.80). A simplified composite score (MOODS, range 0-8) comprised male sex (1 point), overweight (BMI: 25-29.9 kg/m², 1 point) or obesity (BMI: ≥30 kg/m², 3 points), diabetes (2 points), and stroke/transient ischemic attack (2 points) had good discrimination on external validation (C-statistic: 0.73; 95% CI: 0.68-0.77). As a rule-out or a rule-in test, a MOODS score of ≤1 had a 100% sensitivity and score of ≥5 had a 96% specificity for detecting significant SDB, respectively.

The MOODS score provides an individualized and accurate probability of significant SDB in patients with AF. MOODS has the potential to aid clinical decision making and allow efficient resource allocation.