Indicators of care are measures of healthcare delivery that help practitioners make clinical decisions and improve consistency of patient care. They include whether patients are taking specific drugs, or if they have completed certain medical tests. Indicators of care are sometimes developed using the opinion of medical experts only, with little scientific evidence to support them. Historically, this has been a problem as some indicators have been implemented without real world testing, leading to inappropriate treatments.
In this project, we aim to evaluate existing and screen for potential indicators of care in atrial fibrillation, which is the most common abnormal heart rhythm condition. Our project consists of two phases. In phase 1 we will test the effectiveness of existing indicators by comparing outcomes of patients receiving the recommended care with those that are not. In phase 2 we will develop a clinical guideline support tool that can be used to screen for potential new indicators. The public benefit of the research is to ensure the indicators of care used in practice are reliable and effective. Although we use atrial fibrillation as a case study, the methods used in our project are generic and can be applied to other medical conditions.
We will conduct a 2-phase case study on indicators of care in non-valvular atrial fibrillation, the most common sustained cardiac arrhythmia.
In phase 1 we evaluate existing consensus-based indicators of care as described in the literature [1, 2]. Patients will be divided into a control group, for which the indicated care under consideration has not been applied, and an intervention group. The average treatment effect of the indicated care on selected outcomes will be estimated using inverse probability of treatment weighting. To validate the method, we aim to show agreement with our data-driven approach to evidence-based indicators of care in the literature.
In phase 2, a tool that can screen for new potential indicators and re-evaluate the benefit of existing indicators will be developed using regularised survival analysis. We stratify patients into risk groups using Charlson Comorbidity Index, CHA2DS2-VASc and HAS-BLED scores, and screen for clinical indicators of care associated with good patient outcomes within each patient group using the regularisation process. The indicators of care correspond to drug therapies, devices, procedures and pathology tests. The Cox proportional hazard model with regularisation will be used to model the time-to-event outcomes with time-varying covariates covering both patient characteristics and interventions.
All-cause mortality
- Ischaemic stroke and systemic embolism
- Major bleeding
William Tong - Chief Investigator - Macquarie University
William Tong - Corresponding Applicant - Macquarie University
Blanca Gallego Luxan - Collaborator - Macquarie University
Enrico Coiera - Collaborator - Macquarie University
Thierry Wendling - Collaborator - Not from an Organisation
William Runciman - Collaborator - University Of South Australia
HES Admitted Patient Care;HES Outpatient;ONS Death Registration Data;Patient Level Index of Multiple Deprivation