Latent Class Analysis to Determine Patients with Advanced Heart Failure at Highest Risk of Poor Outcomes (RP314)

Abstract

•Examine the relationship between readily available clinical data and poor outcomes such as multiple hospitalization and death for patients with advanced heart failure.•Discuss the use of latent class analytic techniques for finding subgroups or phenotypes within a group of subjects. Predicting prognosis for patients with advanced heart failure (HF) is difficult. Most prediction models don't use regularly captured clinical data. To find a phenotype of patients with advanced HF that might be associated with poor outcomes. Latent class analysis was conducted using secondary data from a 6-site cluster RCT of an intervention to improve communication with advanced HF patients who had implantable cardioverter defibrillators (ICDs). Nine baseline measures were included: hospitalizations in the year before enrollment, HF severity (NYHA class), HF etiology, ICD indication (primary/secondary), functional status (ADLs), symptom burden (>2), comorbidities (>2), depression, and anxiety. Model fit criteria were compared for models testing 1 to 5 latent classes. Two groups of patients emerged: Class1 - “Healthier” - N=352 (62.5%) and Class2 - “More Sick” N=211 (37.5%). Class1 was more likely to be male (76% vs. 61%, p<0.001), married (58% vs. 50%, p<0.05), ischemic (48% vs. 39%, p<0.05) and less severe HF (NYHA class I or II) (12% vs. 2%, p<0.001), and to report no hospitalizations in the prior year (21% vs. 8%, p<0.001). Class2 was more likely to report ADL difficulty (93% vs. 26%, p<0.001), symptom burden (100% vs. 51%, p<0.001), comorbidities (86% vs. 74%, p<0.002), depression (28% vs. 1%, p<0.001), anxiety (30% vs. <1%, p<0.001), and have Medicaid (32% vs. 24%, p<0.05). Differences were not detected for age, race, ethnicity, or education. Patients in Class2 reported more hospitalizations (2.8 vs. 2.2, p<0.02) and were more likely to die (26% vs. 17%, p<0.01). We found a phenotype of advanced HF patients who were more likely to experience multiple hospitalizations and death based on readily available clinical data.