Association between routinely reported symptoms and 3 month hospital admission and mortality risk: a landmark analysis investigation in 86,000 individuals with heart failure in the UK

BackgroundWe investigated symptoms reported before and after heart failure (HF) diagnosis and their associations with 3-month hospitalisation and mortality. ObjectivesTo examine associations between symptoms recorded in primary care and short- term hospitalisation and mortality in HF patients. DesignLandmark analysis using Royston-Parmar survival models at baseline (diagnosis), 6- and 12 months post-diagnosis. SettingPrimary care database (CPRD) linked to hospital and mortality data (1998-2020). ParticipantsAdults (>40 years) with a first HF diagnosis. ExposuresShortness of breath (SOB), ankle swelling, oedema, fatigue, chest pain, depression, and anxiety in the 3 months before diagnosis and at 6 and 12 months. Outcomes3-month all-cause hospitalisation and mortality; secondary outcomes included HF and non-cardiovascular hospitalisation. ResultsAmong 86,882 HF patients (62,742 and 54,555 surviving to 6 and 12 months, respectively), symptom associations varied by timepoint. At diagnosis, depression had the highest risk for all-cause hospitalisation (HR: 1.26; 95% CI 1.15, 1.39) and SOB for HF hospitalisation (1.18; 1.12, 1.26). At 6 months, depression was most associated with all-cause hospitalisation (1.46; 1.25, 1.70), and ankle swelling with mortality (1.49; 1.14, 1.94). At 12 months, SOB had the highest risk for HF hospitalisation (1.99; 1.68, 2.35). ConclusionsSymptoms persisted and were more prominent at 6 and 12 months post- diagnosis than at diagnosis. Strengths and limitationsO_LIThis study used a large, nationally representative cohort from the Clinical Practice Research Datalink (CPRD), enhancing the generalisability of findings to the broader UK heart failure population. C_LIO_LIA dynamic prediction approach (landmark analysis) was employed to account for the time-varying and time-dependent nature of symptoms and covariates, addressing key limitations of static prognostic models. C_LIO_LIRoutinely recorded primary care data were used to capture a broad range of HF- specific and non-specific symptoms, sociodemographic factors, treatments, and comorbidities across multiple clinically relevant timepoints. C_LIO_LISymptoms and covariates were updated at each landmark, allowing for better reflection of patient status over time; however, landmark intervals were selected a priori and may not fully capture individual-level variation. C_LIO_LIKey clinical markers of heart failure severity (e.g., ejection fraction, NYHA class, natriuretic peptides) were not available, which may limit the assessment of symptom relevance across different HF phenotypes. C_LI