Author Affiliations: Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Ioannina, Greece (Drs Siontis and Tzoulaki); Department of Epidemiology and Biostatistics, Imperial College of Medicine, London, England (Dr Tzoulaki); Stanford Prevention Research Center, Department of Medicine, and Department of Health Research and Policy, Stanford University School of Medicine, Stanford, California (Dr Ioannidis); and Department of Statistics, Stanford University School of Humanities and Sciences, Stanford (Dr Ioannidis).
Pilotto et al comment on the Multidimensional Prognostic Index (MPI), a model previously derived and validated from the same investigators, in light of our review on risk-predictive tools for mortality.1 The authors question why 2 validation studies on MPI published during 20092,3 were not captured by our empirical evaluation. The 2 studies were not identified because they used the keyword “ROC curve” rather than the terms “AUC or area under the curve,” which we used in our search. Addition of the “ROC curve” term in our search algorithm would provide 169 additional abstracts to screen including the 2 studies on MPI suggested by Pilotto et al. On the basis of pilot sampling, we estimate that approximately 15 new eligible articles would have been identified, if the search strategy were revised to include the “ROC curve” term as well. This might be a weakness in our current study methodology; however, we do not expect additional articles to be systematically different to the ones included in our review and thus do not expect to alter our conclusions. Our search did not aim to identify all studies on predictive tools for mortality, but to generate a large enough representative sample that would help in understanding the status of this field. This is why we already restricted searches to a single year of publication, since searches without restrictions may have yielded several thousands of articles4 (testimony to the highly prolific nature and low translational efficiency of this literature) without improving much the informativeness of our project. We agree with the authors that multivariable information might be important in predicting future mortality, but our data have shown that most of the available tools have modest accuracy to predict death and high heterogeneity between populations and settings.
Siontis GCM, Tzoulaki I, Ioannidis JPA. A Multidimensional Prognostic Index in Common Conditions Leading to Death in Older Patients—Reply. Arch Intern Med. 2012;172(7):595–596. doi:10.1001/archinternmed.2012.55
Customize your JAMA Network experience by selecting one or more topics from the list below.
Create a personal account or sign in to: