Dementia is a neurodegenerative condition that mostly affects older adults. Given the inevitably ageing population and limited health care resources, the need for reliable and cost-effective dementia interventions are an increasingly important health and social care challenge. As dementia progresses, people with dementia face specific challenges associated with cognitive deterioration which can result in difficulties with self-reporting health status, meaning that, often, there is an increased reliance upon proxy (i.e., formal or informal caregiver) reports for health-related quality of life (HRQoL) assessments. We know that self-report and proxy report often result in different assessments of the health of a person with dementia (PwD), but it is less clear which of the reports is most suitable, whether this differs according to the stage of dementia, or if some combination of the two should be used. Ultimately, these reports are used in assessments of cost effectiveness of heath technologies which in turn are used to inform pricing and policy decisions. Recognised discrepancies between PwD and proxy reports highlight the need for an integrated approach in estimating QALYs for dementia populations.
This thesis aims to analyse differences between HRQoL reports from the PwD and their proxy, with the overarching goal of determining which respondent’s data are optimal for utility score estimations. The thesis includes a review of existing evidence on self and proxy HRQoL assessment, aiming to define the key principles already established in this field. A systematic review is undertaken to assess the psychometric properties of EQ-5D-5L in the context of dementia, aiming to build upon prior findings suggesting the suitability of EQ-5D for dementia evaluation, albeit primarily based on EQ-5D-3L data. Additionally, dementia symptom measures are identified, and a systematic review uses the identified measures to assess convergent validity by contrasting them with EQ-5D's dimensions. The established relationships are used in subsequent analysis chapters. Three distinct dementia trial datasets that collected both PwD and proxy-reported EQ-5D are analysed through psychometric techniques to determine the optimal source of dimension-level data for utility estimations. Additionally, response mapping methods are applied to predict either respondent's data in scenarios where their reports are unavailable.
Key findings can inform data collection decisions and methodologies for future dementia studies. Patient and public involvement and engagement exercises (PPIE) conducted during the PhD reveal challenges associated with reporting specific EQ-5D dimensions, mirroring the subsequent data analysis findings, which demonstrate floor/ceiling effects and missing observations. Proxy reports are found to closely align with symptom measure scores, particularly for "observable" dimensions such as "mobility" and "self-care". Psychometric analyses suggest that these dimensions are best reported by proxies. However, there is insufficient evidence to deviate from PwD reports for the more "non-observable", personally experienced EQ-5D dimensions. Therefore, the evidence suggests that the ideal source for reporting HRQoL for people with dementia may be a combination of PwD and proxy reported EQ-5D dimensions.
In many clinical studies, EQ-5D responses are only obtained from either the PwD or the proxy. Statistical mapping methods are applied to PwD and proxy data by EQ-5D dimension to allow analysts to estimate the full target EQ-5D in such situations. This approach potentially offers a more accurate reflection of dementia experience by integrating both respondent’s reports when estimating HRQoL scores.
The findings also highlight the importance of considering the type of proxy in different residential settings. Staff proxies are deemed more favourable for institutionalised dementia populations. Pragmatic aspects such as instrument administration are appraised for optimising data quality and usability for analyses. This thesis makes a novel contribution to understanding the complexities of HRQoL assessment in dementia populations for use in economic evaluations. A methodologically robust way to integrate PwD and proxy reports via statistical mapping is identified which provides a more comprehensive perspective and a mechanism to overcoming the PwD-proxy gap. This approach provides a valuable tool for future researchers in navigating the challenges inherent in HRQoL research in the context of dementia.
