[Skip to Navigation]
Nov 12, 2012

Good Prediction Rules Are Hard to Find: Did We Succeed?

Author Affiliations

Author Affiliation: Los Angeles County Department of Health Services, Los Angeles, California.

Arch Intern Med. 2012;172(20):1536. doi:10.1001/2013.jamainternmed.120

In a recent commentary, Grady and Berkowitz1 explained why so few of the prediction rules that are submitted to the Archives are ultimately accepted for publication. They reviewed a list of criteria that prediction rules should fulfill to be useful in clinical practice and note that many do not make the cut.

We believe that the PLAN score (a clinical prediction rule comprising preadmission comorbidities, level of consciousness, age, and neurologic deficit) for ischemic stroke,2 while not perfect, will be of use to clinicians. As Grady and Berkowitz noted, prediction rules should use data from patients who are representative of those for whom the rule will be used. The PLAN prediction rule for ischemic stroke was developed on the basis of data from consecutive patients from the Registry of the Canadian Stoke Network.

Add or change institution