What is it about?
A letter published in the Journal of Magnetic Resonance Imaging discusses standardizing the predictive values reported in diagnostic imaging research. The author argues this approach makes study findings more useful to clinicians treating individual patients. He explains that diagnostic tests are often used when the diagnosis is unclear, meaning the pre-test probability of disease is intermediate. Therefore, predictive values adjusted to a standardized 50% disease prevalence provide the most meaningful information to clinicians. The author suggests researchers present raw predictive values along with values standardized to 25%, 50% and 75% prevalences. This accounts for variability in disease prevalence between patients.
Featured Image
Photo by Accuray on Unsplash
Why is it important?
This letter highlights an important issue in applying diagnostic research findings to clinical practice. Study results reflect the population enrolled, but disease prevalence often differs greatly between individual patients. Standardizing predictive values can help clinicians determine the clinical utility of a test for their specific patient. This concept has broad relevance for translating research into evidence-based patient care.
Perspectives
As a practicing physician, I find it challenging to interpret diagnostic study results for my individual patients. A test's predictive value depends heavily on disease prevalence within the study sample. However, prevalence varies widely between patients based on risk factors and clinical presentation. When I order further testing, it's usually when the diagnosis is unclear and pre-test probability is intermediate. In these situations, predictive values standardized to a prevalence of 50% are most useful to me. I appreciate when researchers take care to present data in a clinically meaningful way. Standardized predictive values help me judiciously apply study findings to optimize my diagnostic approach for each patient. This method accounts for the nuances of clinical medicine that are absent from broad population data. I hope more researchers will consider adopting this practice.
Thomas F Heston MD
University of Washington
Read the Original
This page is a summary of: Standardized predictive values, Journal of Magnetic Resonance Imaging, January 2014, Wiley,
DOI: 10.1002/jmri.24564.
You can read the full text:
Resources
Contributors
The following have contributed to this page
