Abstract
BackgroundNHS England (NHSE), the Office for Health Improvement and Disparities (OHID), the World Health Organisation (WHO), and clinical communities require improvements in data collection and use for clinical insight and decision-making for vulnerable groups. Naturalistic healthcare datasets are typically complex and messy, vary in quality, and present analytical challenges. Clinical teams and practitioners may be unaware of these challenges or how to address them, and subsequent analysis risks producing erroneous and/or spurious results, incorrect conclusions, and poor decision-making. This thesis explores: i) To what extent can complex and messy data routinely collected from vulnerable groups be confidently utilised for clinical insight and decision-making? ii) What is the impact of applying advanced statistical analysis to harmonised complex and messy data for vulnerable groups?
Methods
I conducted a critical appraisal of nine selected publications to explore these research questions, focusing on three vulnerable groups: those with severe mental illness; people who use drugs (PWUD); and men who have sex with men (MSM). The groups are linked by their mental health difficulties or high risk of poor mental health, and because they are supported by specialist community-based services in the UK. Included publications demonstrate use of different data types (e.g. electronic health records, surveillance, and clinical trials) and advanced statistical analyses (e.g. multiple imputation).
Findings
My findings are encapsulated by three fundamental themes which demonstrate my contributions to new knowledge in clinical practice: i) improving service engagement, treatment efficiency, and health outcomes - demonstrating how community-based healthcare and non-standard sampling methods can improve engagement, treatment efficiency, and outcomes; ii) translating research findings to a naturalistic setting and widening access to psychological therapies – demonstrating impact of research translation to naturalistic settings, widening access to psychological therapies, and dealing with messy data from routine clinical practice; iii) using assessments of clinically meaningful improvements to make decisions about whether to deliver novel interventions in vulnerable groups - highlighting how clinically meaningful improvement assessments from randomised controlled trials (RCTs) can inform decisions about appropriateness of novel interventions for vulnerable groups in naturalistic settings.
Conclusions
If deemed of good quality according to the eight-domain framework proposed in this thesis (i.e. efficient, accurate, valid, reliable, complete, relevant, granular, and/or timely), complex and messy data acquired from vulnerable groups can be confidently used by clinicians and health practitioners for clinical insight and decision-making. This includes evidence of acceptability of a novel treatment or indicating how interventions should be tailored. Advanced statistical analyses, such as multiple imputation and multilevel modelling, can successfully be combined with data harmonisation methods to: understand individual needs better; generate evidence of patient benefits of novel interventions and care models; identify ways to improve patient journey experiences; and ultimately positively impact health. Furthermore, due to the frequent occurrence of data missingness in naturalistic clinical datasets, the significant role that advanced statistical analysis (in particular multiple imputation) plays in enhancing data quality should not be ignored.
| Date of Award | Jul 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Nigel Sherriff (Supervisor), Kathleen Galvin (Supervisor) & Jorg Huber (Supervisor) |