Abstract 13353: Sleep Health Phenotypes Predict Symptoms, Cognitive Function, and Adverse Events in Patients With Heart Failure

Introduction: Poor sleep contributes to outcomes in heart failure (HF). Phenotypes reflecting multi-dimensional sleep health may better predict outcomes and support more precise interventions than single measures, but few studies have used this approach.

Aims: 1) Describe phenotypes of sleep health and 2) Evaluate whether sleep health phenotypes predict symptoms, cognitive function, and time to first hospitalization and emergency department (ED) visits among HF patients.

Methods: We performed secondary analysis of a randomized controlled trial of cognitive behavioral therapy for insomnia compared with HF self-management education with HF patients. We measured sleep [rest-activity rhythms, sleep duration, and efficiency (2-week wrist actigraphy/follow-ups); insomnia; CPAP use; and daytime sleepiness (Epworth Scale)]; perceived stress; fatigue; anxiety; depression; pain; and cognitive ability (Patient-Reported Outcomes Measurement System: PROMIS) at baseline and 3, 6, and 12 months post-intervention and tracked hospitalizations and ED visits. We used K-means clustering to identify sleep phenotypes and generalized linear mixed models and Cox models to predict outcomes, adjusted for demographic factors.

Results: Among 166 participants (M age = 63.2 (SD = 12.8) years; 57% female; 23% NYHA Class 3/4 HF; 35% EF less than 45%), there were 4 phenotypes [poor sleep (14%); short sleep/high sleep efficiency (39%); long sleep/low sleep efficiency (25%); and healthy sleep (21%)] at baseline. The healthy sleep group was older and had lower BMIs and less comorbidity and had better outcomes compared to all other phenotypes. In particular, this group had better fatigue (-2.43±0.87), pain (-1.86±1.05), anxiety (-1.50±0.89), stress (-1.46±0.69), and cognitive ability (1.76±0.80) than the poor sleep group. The poor sleep group had shorter time to adverse events compared to the other phenotypes (Hazard ratios of 0.41 to 0.56 for hospitalizations and 0.35 to 0.60 for ED visits).

Conclusions: Multidimensional aspects of sleep predict HF outcomes and may improve on prediction with single measures. Future tailored sleep interventions focused on phenotypes may more precisely improve outcomes among HF patients than traditional approaches.