Assessment and comparison of prognostic classification schemes for survival data

Graf, Erika; Schmoor, Claudia; Sauerbrei, Willi; Schumacher, Martin

doi:10.1002/(sici)1097-0258(19990915/30)18:17/18<2529::aid-sim274>3.3.co;2-x

Cited by 226 publications

(354 citation statements)

References 0 publications

Supporting

Mentioning

350

Contrasting

Unclassified

Order By: Relevance

“…However, when the purpose is to compare the performance of different, potentially misspecified, models it is preferable to use an approach that does not depend on the models which are compared. Such 'model-free' methods have been described in Heagerty et al (2000); Uno et al (2007) for ROC analysis and in Graf et al (1999); Gerds and Schumacher (2006) for the Brier score. We refer to these articles for details of computation.…”

Section: Censored Datamentioning

confidence: 99%

“…This was studied in Graf et al (1999); Uno et al (2007); Gerds and Schumacher (2006) for the performance measures discussed in this article. The IPCW approach requires an estimate of the conditional censoring survival function.…”

Section: Censored Datamentioning

confidence: 99%

“…There is however an important special case: If it is reasonable to assume that the censoring distribution is independent of the predictors, then one can use the KaplanMeier estimator for the censoring distribution in (15). This approach was proposed in Graf et al (1999). However, as pointed out in Van der Laan and Robins (2003) this approach is inefficient.…”

Section: Censored Datamentioning

confidence: 99%

See 2 more Smart Citations

The Performance of Risk Prediction Models

Gerds¹,

Cai²,

2008

Self Cite

View full text Add to dashboard Cite

SummaryFor medical decision making and patient information, predictions of future status variables play an important role. Risk prediction models can be derived with many different statistical approaches. To compare them, measures of predictive performance are derived from ROC methodology and from probability forecasting theory. These tools can be applied to assess single markers, multivariable regression models and complex model selection algorithms. This article provides a systematic review of the modern way of assessing risk prediction models. Particular attention is put on proper benchmarks and resampling techniques that are important for the interpretation of measured performance. All methods are illustrated with data from a clinical study in head and neck cancer patients.

show abstract

Section: Censored Datamentioning

confidence: 99%

Section: Censored Datamentioning

confidence: 99%

Section: Censored Datamentioning

confidence: 99%

See 1 more Smart Citation

The Performance of Risk Prediction Models

Gerds¹,

Cai²,

2008

Self Cite

View full text Add to dashboard Cite

show abstract

“…We follow this line of thinking and adapt the prediction error proposal of Graf et al (1999) and Gerds and Schumacher (2006) to handle models with more than one possible event type. We focus on predictions made in terms of predicted probabilities, as it is known that time point predictions are usually of poor value (Henderson et al, 2001).…”

Section: Introductionmentioning

confidence: 99%

“…However, research into methods for the evaluation of the prognostic potential of these models is still needed, as most proposed methods measure either discrimination or calibration of models, but do not examine both simultaneously. We adapt the prediction error proposal of Graf et al (1999) and Gerds and Schumacher (2006) to handle models with more than one possible event type and introduce a consistent estimator. A simulation study investigating the behaviour of the estimator in small sample size situations and for different levels of censoring together with a real data application follows, highlighting the usefulness of the proposed approach for quantifying effects of model misspecification in summary models for a competing risks setting.…”

mentioning

confidence: 99%

Quantifying the predictive accuracy of time-to-event models in the presence of competing risks

et al. 2011

Self Cite

View full text Add to dashboard Cite

The development of prognostic time to event models is an area of active research, notably in the fields of cardiology, oncology and intensive care medicine. These models try to find a link between patient covariates and an event at a later time point, and are used for example for therapy assignment, risk stratification or interhospital quality assurance. Increasingly, interest is not only in prognostic models with one possible type of event (e.g. death), but in models that distinguish between several competing endpoints. However, research into methods for the evaluation of the prognostic potential of these models is still needed, as most proposed methods measure either discrimination or calibration of models, but do not examine both simultaneously. We adapt the prediction error proposal of Graf et al. (1999) and Gerds and Schumacher (2006) to handle models with more than one possible event type and introduce a consistent estimator. A simulation study investigating the behaviour of the estimator in small sample size situations and for different levels of censoring together with a real data application follows, highlighting the usefulness of the proposed approach for quantifying effects of model misspecification in summary models for a competing risks setting.

show abstract