Background: Individual symptoms are predictors of mortality in adults with heart failure (HF). However, symptom clusters may be more predictive of future risks than isolated symptoms, yet research on these symptom clusters in older adults with HF is limited. Aims: To explore the extent to which symptom cluster profiles predict all-cause mortality among community-dwelling older adults with HF, while adjusting for known risk factors. Methods: We conducted a secondary analysis of data of the 2008 to 2016 surveys of the U.S. Health and Retirement Study using latent class analysis to identify baseline cluster profiles, and survival analysis for time to death with Cox proportional hazard (Cox PH) models, and Kaplan Meier survival analyses. Results: We included 684 participants [mean age=74.9, (SD=10) years, 56.6% female, 16.2% Black/African American]. Three baseline symptom cluster profiles were identified: high-burden (pain, shortness of breath, fatigue, swelling, depressive symptoms, dizziness), low-burden, and cardiopulmonary-depressive (shortness of breath, pain, and dizziness). The estimated median time-to-death was 71 (95% CI=64, 79) months. Of the 364 participants in the cardiopulmonary-depressive profile, 240 (65.9%) died (median time= 65 months, 95% CI= 55, 73) compared to those in the low-burden profile, of which 49% died. Approximately 61% in the high burden profile died (median time=67 months, 95% CI= 51, 90). A significant difference in survival times between the 3 cluster profiles was found (Log-rank= 9.13, p = 0.01). In the adjusted Cox PH model, participants in the high symptom burden and respiratory-depressive distress profiles had adjusted HR of 1.48 (95% CI=1.15, 1.94) and 1.44 (95% CI=1.14, 1.80) compared to those in the low burden symptom cluster profile, after controlling for age, gender, smoking, and comorbidities. Conclusions: The identified symptom cluster profiles predict 8-year all-cause mortality in older adults with HF, after controlling for known risk factors. An evaluation of symptom clusters, rather than individual symptoms, may provide additive prognostic information.
Read full abstract