Smooth time-dependent receiver operating characteristic curve estimators

Martínez–Camblor, Pablo; Pardo–Fernández, Juan Carlos

doi:10.1177/0962280217740786

Cited by 21 publications

(30 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These scenarios are very similar to those of Martínez‐Camblor and Pardo‐Fernández 28 . The values of p are set to be 0.25 and 0.50.…”

Section: Simulation Studymentioning

confidence: 91%

“…In this section, we present the results of a simulation study conducted to compare the finite sample performance of the proposed boundary corrected ROC curve estimator with the proposed empirical (nonsmoothed) estimator and the MP method of Martínez-Camblor and Pardo-Fernández. 28 Table 5 (Table 6) shows the MIB and the MISE of the first (second) scenario obtained for different sample sizes, censoring rates, p values, and time horizons. As can be seen from these tables, the MP method has a notably larger MIB and MISE than both the empirical (nonsmoothed) and the boundary corrected smooth estimators proposed in this work; see Equations (4) and (8), respectively.…”

Section: Comparing the Proposed Methods With The Competitorsmentioning

confidence: 99%

“…Note that, in contrast to cross‐sectional studies, where there is an abundance literature about smoothing ROC curves, see the recent paper of Pulit 27 and the references given therein, this problem is not well studied in the context of survival analysis. In fact, to the best of our knowledge, the only related work is the recent paper of Martínez‐Camblor and Pardo‐Fernández 28 . These authors proposed a smooth time‐dependent ROC curve estimators for both cumulative sensitivity/dynamic specificity and incident sensitivity/dynamic specificity.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Smoothed time‐dependent receiver operating characteristic curve for right censored survival data

Beyene

Ghouch

2020

Statistics in Medicine

View full text Add to dashboard Cite

The prediction reliability is of primary concern in many clinical studies when the objective is to develop new predictive models or improve existing risk scores. In fact, before using a model in any clinical decision making, it is very important to check its ability to discriminate between subjects who are at risk of, for example, developing certain disease in a near future from those who will not. To that end, the time-dependent receiver operating characteristic (ROC) curve is the most commonly used method in practice. Several approaches have been proposed in the literature to estimate the ROC nonparametrically in the context of survival data. But, except one recent approach, all the existing methods provide a nonsmooth ROC estimator whereas, by definition, the ROC curve is smooth. In this article we propose and study a new nonparametric smooth ROC estimator based on a weighted kernel smoother. More precisely, our approach relies on a well-known kernel method used to estimate cumulative distribution functions of random variables with bounded supports. We derived some asymptotic properties for the proposed estimator. As bandwidth is the main parameter to be set, we present and study different methods to appropriately select one. A simulation study is conducted, under different scenarios, to prove the consistency of the proposed method and to compare its finite sample performance with a competitor. The results show that the proposed method performs better and appear to be quite robust to bandwidth choice. As for inference purposes, our results also reveal the good performances of a proposed nonparametric bootstrap procedure. Furthermore, we illustrate the method using a real data example.

show abstract

“…These scenarios are very similar to those of Martínez‐Camblor and Pardo‐Fernández 28 . The values of p are set to be 0.25 and 0.50.…”

Section: Simulation Studymentioning

confidence: 91%

Section: Comparing the Proposed Methods With The Competitorsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Smoothed time‐dependent receiver operating characteristic curve for right censored survival data

Beyene

Ghouch

2020

Statistics in Medicine

View full text Add to dashboard Cite

show abstract

“…The clinical outcomes are time-dependent, for example a healthy person may have disease over the follow-up time. Hence, we applied incident/dynamic time-dependent area under the receiver operating characteristic curves (AUCs) that accounts for time in order to compare the predictive abilities of FEV1 expressions and their respective methods of classification of COPD severity to predict all-cause mortality, respiratory mortality, cardiovascular mortality, COPD hospitalization, and pneumonia hospitalization (31)(32)(33)(34). For cause-specific mortality and hospitalization, AUCs accounting for competing risks were calculated (33).…”

Section: Discussionmentioning

confidence: 99%

Spirometric classifications of COPD severity as predictive markers for clinical outcomes: the HUNT Study

Bhatta

Leivseth

Mai

et al. 2020

Preprint

View full text Add to dashboard Cite

Rationale GOLD grades based on percent-predicted FEV1 poorly predicts mortality. Studies have recommended alternative expressions of FEV1 for the classification of COPD severity and they warrant investigation. Objective To compare the predictive abilities of ppFEV1 (ppFEV1 quartiles, GOLD grades, ATS/ERS grades), FEV1 z-score (FEV1 z-score quartiles, FEV1 z-score grades), FEV1.Ht-2 (FEV1.Ht-2 quartiles, FEV1.Ht-2 grades), FEV1.Ht-3 (FEV1.Ht-3 quartiles), and FEV1Q (FEV1Q quartiles) to predict clinical outcomes. Methods People aged ≥40 years with COPD (n=890) who participated in the HUNT Study (1995-1997) were followed for 5 years (short-term) and up to 20.4 years (long-term). Survival analysis and time-dependent area under curve (AUC) were used to compare the predictive abilities. A regression tree approach was applied to obtain optimal cut-offs of different expressions of FEV1. The UK Biobank (n=6495) was used as a replication cohort with a 5-year follow-up. Measurements and Main Results As a continuous variable, FEV1Q had the highest AUCs for all-cause mortality (short-term 70.2, long-term 68.3), respiratory mortality (short-term 68.4, long-term 67.7), cardiovascular mortality (short-term 63.1, long-term 62.3), COPD hospitalization (short-term 71.3, long-term 70.9), and pneumonia hospitalization (short-term 67.8, long-term 66.6), followed by FEV1.Ht-2 or FEV1.Ht-3. Generally, similar results were observed for FEV1Q quartiles. The optimal cut-offs of FEV1Q had higher AUCs compared to GOLD grades for predicting short-term and long-term clinical outcomes. Similar results were found in UK Biobank. Conclusions FEV1Q best predicted the clinical outcomes and could improve the classification of COPD severity.

show abstract

“…(). Many other methods to estimate the time‐dependent AUC exist in the literature; see, for example, Martínez‐Camblor and Pardo‐Fernández () and the references given therein. For a comprehensive summary of the time‐dependent ROC curve estimation methods, the reader is referred to Blanche et al.…”

Section: Introductionmentioning

confidence: 99%

On the validity of time‐dependent AUC estimation in the presence of cure fraction

2019

View full text Add to dashboard Cite

During the last decades, several approaches have been proposed to estimate the timedependent area under the receiver operating characteristic curve (AUC) of risk tools derived from survival data. The validity of these estimators relies on some regularity assumptions among which a survival function being proper. In practice, this assumption is not always satisfied because a fraction of the population may not be susceptible to experience the event of interest even for long follow-up. Studying the sensitivity of the proposed estimators to the violation of this assumption is of substantial interest. In this paper, we investigate the performance of a nonparametric simple estimator, developed for classical survival data, in the case when the population exhibits a cure fraction. Motivated from the current practice of deriving risk tools in oncology and cardiovascular disease prevention, we also assess the loss, in terms of predictive performance, when deriving risk tools from survival models that do not acknowledge the presence of cure. The simulation results show that the investigated method is valid even under the presence of cure. They also show that risk tools derived from survival models that ignore the presence of cure have smaller AUC compared to those derived from survival models that acknowledge the presence of cure. This was also attested with a real data analysis from a breast cancer study. K E Y W O R D SCox model, mixture cure models, promotion time models, risk tools, time-dependent AUC 1430

show abstract

Smooth time-dependent receiver operating characteristic curve estimators

Cited by 21 publications

References 32 publications

Smoothed time‐dependent receiver operating characteristic curve for right censored survival data

Smoothed time‐dependent receiver operating characteristic curve for right censored survival data

Spirometric classifications of COPD severity as predictive markers for clinical outcomes: the HUNT Study

On the validity of time‐dependent AUC estimation in the presence of cure fraction

Contact Info

Product

Resources

About