Longitudinal trajectories of health indicators using real world data

Older people with complex health needs are often excluded from clinical trials, primarily due to factors such as age, multimorbidity, and polypharmacy. However, they represent a significant portion of healthcare resource consumption and the use of newly authorised medications. Existing guidelines for identifying and treating this population often rely on using cross-sectional values and providing guidance on treating individual conditions rather than addressing the complexities of multimorbidity and treatment combinations. In this thesis, I propose applying novel approaches for identifying and characterising older people with complex health needs, using different health indicators and study designs on real world data. The definitions and cohorts established in this work have the potential to inform decisions for identifying, managing and treating older people with complex health needs. In the first project, I conducted a cross-sectional analysis to identify three cohorts of older people with high levels of frailty, polypharmacy, or unplanned hospital admissions. Patients in any of these cohorts had high comorbidity burden and preventive therapy use. Although there was considerable overlap between these cohorts, many patients only belonged to one of the three cohorts. This indicates that these health markers are intersectional and complementary to each other. Frailty and polypharmacy are cumulative conditions that take years to develop, making cross-sectional cohorts unable to describe their progression over time. In projects two and three, I modelled frailty and polypharmacy in older people over 4-5 years of follow-up. I identified subgroups with distinct frailty or polypharmacy trajectories, which demonstrated different association levels with the risk of mortality. Most of the population belonged to the low-steady/slow (healthy) subgroup. However, important subgroups emerging from these studies started from a seemingly healthy state, deteriorated rapidly over the study follow-up and had the highest mortality risks, indicating their need for more healthcare resources and monitoring. The subgroups were identified in a UK primary care database and then externally validated in two independent national and international databases. They demonstrated generalisability with good external validity, similar trajectories and clinical characteristics. Previous evidence reported that frailty and polypharmacy could start from middle age, and some of the identified subgroups of older people started from elevated or intermediate levels of frailty and polypharmacy. To understand these health markers’ progression from early on, I modelled polypharmacy over time in middle-aged people in the fourth project. I identified three subgroups with distinct polypharmacy trajectories and associated mortality risks. I found that those with the fastest polypharmacy trajectory had the highest mortality risk, followed by those starting at the highest polypharmacy baseline values. Those patients are likely to continue progressing and end up in one of the non-healthy subgroups at older age. My research demonstrated that monitoring trajectories of frailty and polypharmacy predicts mortality better than cross-sectional values. The identified subgroups were generalisable and had distinctive clinical characteristics. Future research can focus on further generalisability of the identified subgroups, and investigate how polypharmacy and frailty progress over longer periods, together and individually.