Investigating the association between symptoms and clinical outcomes in patients with heart failure and multiple long term conditions: an electronic health record study using dynamic modelling

Background: Symptoms as prognostic factors for hospitalisation and mortality in heart failure patients show conflicting findings. There is no systematic review for typical heart failure (HF) symptoms (dyspnoea, fatigue and peripheral oedema) and their relationship to hospitalisation and mortality. Furthermore, evidence of symptoms as prognostic factors is overwhelmingly at one time point and usually in secondary care, although most patients with HF are managed in a primary care setting.

Methods: The research was comprised of four phases: a systematic review synthesising evidence on HF symptoms as prognostic factors; investigating the association between time-varying symptoms and outcomes (first hospitalisation and mortality) using risk-set sampling in a nested case-control study design using the Clinical Practice Research Datalink, survival analysis examining symptom associations at HF diagnosis and rates of adverse outcomes over 12 months and dynamic modelling using landmarking at diagnosis, 6- and 12-months to assess symptom associations to outcomes over a three-month time-horizon.

Results: The systematic review showed that overall, symptoms in HF patients showed unclear findings for all-cause hospitalisation and death. Furthermore, there was limited data from primary/real-world evidence. The risk-set sampling study showed a strong association between time-varying symptoms and outcomes. Using a cohort analysis to investigate theassociation between symptoms present at diagnosis and 12-month outcomes showed (i) mixed and often non-significant associations and (ii) symptom associations often depended on follow-up time. Finally, using dynamic modelling showed differential associations between symptoms and 3-month outcomes across the different landmark points. Compared to diagnosis, risk associations increased at 6 and 12 months.

Conclusion: This thesis contributed to the evidence regarding the association between routinely reported symptoms and outcomes in patients with HF. Key findings include: (i) symptoms are strongly associated with hospitalisation and mortality and (ii) symptoms associations are influenced by the time of their presentation and duration of follow-up.