This research project is in progress. PCORI will post its findings here within 90 days after our final review is complete. In the meantime, results have been published in peer-reviewed journals, as listed below.
Objective: Registries, such as the Society for Vascular Surgery’s Vascular Quality Initiative (VQI) and administrative billing records, are an important source of data on patients’ treatment and health outcomes. They involve large numbers of patients and are reflective of real-world complexities surrounding treatment. The primary concern with such sources of data is that the treatments patients select may be determined by data not included in the registry, such as details of a patient’s medical information, that also affect outcomes. Such data are known as unmeasured confounding variables and, if ignored, are likely to bias results. The only way of overcoming unmeasured confounders in the VQI and other patient registries is to identify a natural experiment by finding a variable that predicts the treatment a patient receives but otherwise does not affect their outcome. Such a variable is an instrumental variable (IV).
Although the time until an event of interest is a common patient outcome, IV methods for these data are underdeveloped. This project will develop IV methods for the Cox model, the most common approach to assessing the comparative effectiveness of treatments using time-to-event data. A second problem we address is the lack of knowledge about the best format of risk information and the best visualization method to use when communicating with patients. To do so, we will conduct a qualitative study of VQI patients.
- Development of IV methodology for the Cox model and confirmation that it yields more accurate risk information for patients
- Development of IV methodology for the extended Cox model in which treatment effects vary over follow-up or pre-specified subpopulations
- Engagement of VQI patients with carotid artery disease on the presentation of risk information from Cox model analyses of VQI patient data
We expect that IV–Cox model analyses will yield more accurate risk information than approaches that ignore unmeasured confounding, and that patient interviews will yield valuable insights into the most effective format of presenting risk information.
Importance to Patients: Patients and their physicians need better evidence to guide choices among diagnostic and therapeutic approaches, which relies on the appropriateness of the methods generating the evidence. Therefore, more accurate information on the comparative effectiveness of treatments will facilitate more informed decision making. Together with patient feedback on formats of presenting risk information, this project will provide the basis for patients to make better treatment decisions.
Role of Patient and Stakeholder Partners:
We worked with two patients with asymptomatic carotid stenosis. Both provided valuable feedback and support for the project. They will join physicians, researchers, and other professionals on the nine-member Patient Engagement and Governance Committee, which will ensure that patients are central to the research and its results.