http://www.biomedcentral.com/bmcmedresmethodol/ The editor's pick of recent articles published by BMC Medical Research Methodology 2013-03-06T00:00:00Z Background A prognostic model should not enter clinical practiceunless it has been demonstrated that it performs a useful role. Externalvalidation denotes evaluation of model performance in a sample independent of that used to develop the model. Unlike for logistic regression models,external validation of Cox models is sparsely treated in the literature.Successful validation of a model means achieving satisfactory discriminationand calibration (prediction accuracy) in the validation sample. Validating Coxmodels is not straightforward because event probabilities are estimatedrelative to an unspecified baseline function.Methods We describe statistical approaches to external validation ofa published Cox model according to the level of published information,specifically (1) the prognostic index only, (2) the prognostic index togetherwith Kaplan-Meier curves for risk groups, and (3) the first two plus thebaseline survival curve (the estimated survival function at the meanprognostic index across the sample). The most challenging task, requiringlevel 3 information, is assessing calibration, for which we suggest a methodof approximating the baseline survival function.Results We apply the methods to two comparable datasets in primarybreast cancer, treating one as derivation and the other as validation sample.Results are presented for discrimination and calibration. We demonstrate plotsof survival probabilities that can assist model evaluation.Conclusions Our validation methods are applicable to a wide range ofprognostic studies and provide researchers with a toolkit for externalvalidation of a published Cox model. http://www.biomedcentral.com/1471-2288/13/33 Patrick Royston Douglas G Altman BMC Medical Research Methodology 2013, 13:33 2013-03-06T00:00:00Z 10.1186/1471-2288-13-33 Principles of Cox model validation <p>Statistical approaches to validate Cox prognostic models, which are often difficult, are applied; the approaches represent valuable tools for researchers wanting to validate published Cox models.</p> /content/figures/1471-2288-13-33-toc.gif BMC Medical Research Methodology 1471-2288 13 33 2013-03-06T00:00:00Z PDF Background: The occurrence of communicable diseases (CD) depends on exposure to contagious persons. The effects of exposure to CD are delayed in time and contagious persons remain contagious for several days during which their contagiousness varies. Moreover when multiple exposures occur, it is difficult to know which exposure is associated with the CD. Methods: A statistical model at the individual level is presented to estimate the risk of CD to patients, in healthcare settings, with multiple observed exposures to other patients and healthcare workers and unobserved exposures to unobserved or unobservable sources. The model explores the delayed effect of observed exposure, of source contagiousness and of unobserved exposure. It was applied to data on influenza-like illness (ILI) among patients in a university hospital during 3 influenza seasons: from 2004 to 2007. Over a total of 138,411 patients-days of follow-up, 64 incident ILI cases were observed among 21,519 patients at risk of ILI. Results: The ILI risk per 10,000 patients-days associated with observed exposure was about 129.1 (95% Credible Interval (CrI): 84.5, 182.9) and was associated at 72% with exposures to patients or healthcare workers 1 day earlier and at 41% with the 1st day of source contagiousness. The ILI risk associated with unobserved exposure was 0.8 (95% CrI: 0.3, 1.6) per 10,000 patients-days in non-epidemic situation in the community and 4.3 (95% CrI: 0.4, 11.0) in epidemic situation. Conclusions: The model could be an interesting epidemiological tool to further assess the relative contributions of observed and unobserved exposures to CD risk in healthcare settings. http://www.biomedcentral.com/1471-2288/13/26 Cécile Payet Nicolas Voirin Philippe Vanhems René Ecochard BMC Medical Research Methodology 2013, 13:26 2013-02-20T00:00:00Z 10.1186/1471-2288-13-26 Assessing risks with multiple exposures <p>A statistical model to overcome the difficulties in assessing risks of communicable diseases with multiple exposures could serve as a useful new tool in the epidemiology of infection.</p> /content/figures/1471-2288-13-26-toc.gif BMC Medical Research Methodology 1471-2288 13 26 2013-02-20T00:00:00Z XML