Intervening with a manualised package to achieve treatment adherence in people with Tuberculosis (impact): feasibility of collecting cost and quality of life data from records and patients

Introduction and Objectives
Compared to the rest of the UK and Western Europe, England has high rates of tuberculosis (TB). Non-adherence to TB treatment can be detrimental for patients’ health and lead to onward infection. The IMPACT study developed a complex manualised intervention including an enhanced TB needs assessment (TNA) for use in routine care within the National Health Service (NHS) to address the demographic, medical, practical, structural, and psychosocial factors contributing to poor adherence. For the health economics workstream of this feasibility study, we aimed to assess the feasibility of capturing useable health economic data in this context, with a view to planning a cost-effectiveness analysis in a future full trial.

Methods
Client Service Receipt Inventory (CSRI) and EQ-5D-5L questionnaires were administered at baseline and two, four, and six months post-baseline to 79 adult study participants from four sites in London to collect information on healthcare resource use and health-related quality of life (HRQoL), respectively. Use of TB-related medications was captured from records at baseline and every follow-up timepoint (two weeks, then monthly from one to six months). Time spent delivering the intervention or standard of care was also captured in the respective arm. Descriptive statistics including levels of missingness were produced for the whole group and by intervention arm. No formal statistical testing was performed due to the small sample size, in accordance with the analysis plan.

Results
Completion rates of CSRI and EQ-5D-5L questionnaires ranged between 71–99% across timepoints, with differing degrees of missingness and data queries required according to the type of information captured. Utility scores could be calculated from EQ-5D-5L responses for HRQoL, and costs could be calculated for the different categories of healthcare resources used by applying unit cost information to the captured resource use information, with input from clinical and other colleagues on the study team.

Conclusions
Our preliminary results suggest that collecting patient-reported and record-based resource use and HRQoL data is feasible in this patient group. Further refinement of the database design would be required to make the collection and analysis of resource use data less labour intensive for the future study.