Abstract
Medication adherence can be defined as ‘the process by which patients take their medications as prescribed’. Suboptimal adherence can have negative impacts on health outcomes, including accelerated disease progression, increased risk of mortality and increased healthcare costs.
A quantitative understanding of medication-taking is required to gauge the extent and magnitude of suboptimal adherence behaviours. This in turn is contingent on measurement methods and outcome metrics that provide an unbiased picture of adherence behaviour. An important aspect of adherence research is therefore the identification and mitigation of biases that may impact the inferences from adherence studies. Guidance to assist researchers in the understanding of bias risk when conducting or reviewing adherence research is not available.
The central aim of this thesis was to identify and understand biases associated with adherence research. This includes facilitating the assessment of bias risk in adherence studies using purpose-designed tools and mitigating biases expected to impact the measurement, analysis and reporting of adherence.
A systematic review assessing the quantitative impact of clinical interventions on medication-taking behaviours in patients with gout found that most studies were at high to serious risk of bias in aspects relating to study design, outcome measurement, and reporting. The use of existing risk of bias tools highlighted the need for purpose-designed bias tools for adherence studies.
A framework was developed to identify and collate sources of bias important in medication adherence research including suggestions about how these biases can be mitigated. The twenty three biases identified were mapped onto the different phases of adherence, the measurement methods and the associated metrics used to summarise adherence outcomes. The framework was intended to inform the future design of adherence studies and the development of risk of bias tools for adherence research.
Biases expected to impact the design and interpretation of adherence studies were used as the basis for the development of risk of bias tools specific for medication adherence studies. Two risk of bias tools were developed; the Risk of Bias tool for Interventional Adherence Studies (RoBIAS) and the Risk of Bias tool for Observational Adherence Studies (RoBOAS). The tools encompassed four Domains relating to; (i) study design, (ii) randomisation (RoBIAS tool) and confounding factors (RoBOAS tool), (iii) adherence outcome measurement and (iv) data analysis. Each domain consisted of detailed items/statements, each mapped to specific biases. The tools were intended to have utility when systematically reviewing published adherence research and to inform the design of future studies.
The RoBIAS and RoBOAS tools were piloted and evaluated using a cohort of twenty expert adherence researchers. Expert opinion allowed for valuable feedback in finalising the developed tools and improving the tool’s utility. A guidance document was developed to aid users of the tools and assist in bias judgement decisions.
The finalised bias tools were applied and showcased in an updated review of adherence interventional studies for patients with gout. There was large heterogeneity observed between the new studies in the measurement methods used, study designs employed and outcome metrics utilised. An informal comparison between the adherence bias tools with the generic bias tools used in the original review highlighted important biases that were detected by the RoBIAS and RoBOAS tools.
The work from this thesis provides insight into the sources of bias associated with adherence research. The novel risk of bias tools for adherence studies make a unique contribution to the field. Understanding sources of bias is critical to enhance the quality and rigour of adherence studies and ultimately, provide an unbiased assessment of medication-taking behaviour.