November 18, 2016: Lumbar Imaging with Reporting of Epidemiology (LIRE)
Jeffrey (Jerry) Jarvik, MD, MPH, Departments of Radiology, Neurological Surgery, Health Services, and the Comparative Effectiveness, Cost, and Outcomes Research Center at the University of Washington
Patrick Heagerty, PhD, Professor and Chair, Department of Biostatistics, and Director of the Center for Biomedical Statistics at the University of Washington
Lumbar Imaging with Reporting of Epidemiology (LIRE): The Beginning of the End (or The End of the Beginning?)
Lumbar Imaging with Reporting of Epidemiology; LIRE; Demonstration Project; Radiology reports; Epidemiology; Prevalence benchmark data; Natural language processing; Spine imaging; Health-related quality of life; Stepped-wedge randomization; Cluster randomization; Electronic health records; EHRs
- The key pragmatic aspects of the LIRE trial are broad inclusion criteria; waiver of consent; centralized IRB review; simple, easily implemented intervention; passive collection of outcomes; and stepped-wedge cluster randomization.
- Stepped-wedge randomization is a form of cluster randomization that involves random and sequential crossover of clusters from control to intervention until all clusters are exposed to the intervention.
- Lumbar spine imaging frequently reveals incidental findings. In the LIRE trial, epidemiologic benchmark data are inserted in lumbar spine imaging reports with the goal of reducing subsequent tests and treatments.
- Report insertion text: “The following findings are so common in normal, pain-free volunteers that, while we report their presence, they must be interpreted with caution and in the context of the clinical situation. Among people between the age of 40 and 60 years who do not have back pain, a plain film x-ray will find that about 8 in 10 have disk degeneration; 6 in 10 have disk height loss. Note that even 3 in 10 means that the finding is quite common in people without back pain.”
- It has been important in the LIRE trial to engage as broad a group of providers (primary care physicians and radiologists) as possible.
Challenges of the LIRE trial include use of multiple information systems, varying EHR implementations, and the merging of datasets.
Were there missing data or loss to follow-up in your data over time? How does the study team account for, and adjust the data analysis for, temporal trends?
What was your experience with using natural language processing (NLP)—and do you think it can be reproduced outside of your set of systems? Would collaborating with NLP experts help to achieve better results?
For More Information
Visit the LIRE Demonstration Project’s webpage.
@PCTGrandRounds, @Collaboratory1, @PCORnetwork