Validation of prediction models in the presence of competing risks: a guide through modern methods

Nan Van Geloven; Daniele Giardiello; Edouard F. Bonneville; Lucy Teece; Chava L. Ramspek; Maarten Van Smeden; Kym I.E. Snell; Ben Van Calster; Maja Pohar-Perme; Richard D. Riley; Hein Putter; Ewout Steyerberg

doi:10.1136/bmj-2021-069249

Validation of prediction models in the presence of competing risks: a guide through modern methods

Nan Van Geloven, Daniele Giardiello, Edouard F. Bonneville, Lucy Teece, Chava L. Ramspek, Maarten Van Smeden, Kym I.E. Snell, Ben Van Calster, Maja Pohar-Perme, Richard D. Riley, Hein Putter, Ewout Steyerberg^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › Academic › peer-review

Abstract

Thorough validation is pivotal for any prediction model before it can be advocated for use in medical practice. For time-to-event outcomes such as breast cancer recurrence, death from other causes is a competing risk. Model performance measures must account for such competing events. In this article, we present a comprehensive yet accessible overview of performance measures for this competing event setting, including the calculation and interpretation of statistical measures for calibration, discrimination, overall prediction error, and clinical usefulness by decision curve analysis. All methods are illustrated for patients with breast cancer, with publicly available data and R code.

Original language	English
Article number	e069249
Journal	The BMJ
Volume	377
DOIs	https://doi.org/10.1136/bmj-2021-069249
Publication status	Published - 24 May 2022

Keywords

Humans
Models, Statistical
Proportional Hazards Models
Risk Assessment
Risk Factors

Access to Document

10.1136/bmj-2021-069249

Cite this

@article{ae595422cb4e47469e0973a3434371ff,

title = "Validation of prediction models in the presence of competing risks: a guide through modern methods",

abstract = "Thorough validation is pivotal for any prediction model before it can be advocated for use in medical practice. For time-to-event outcomes such as breast cancer recurrence, death from other causes is a competing risk. Model performance measures must account for such competing events. In this article, we present a comprehensive yet accessible overview of performance measures for this competing event setting, including the calculation and interpretation of statistical measures for calibration, discrimination, overall prediction error, and clinical usefulness by decision curve analysis. All methods are illustrated for patients with breast cancer, with publicly available data and R code. ",

keywords = "Humans, Models, Statistical, Proportional Hazards Models, Risk Assessment, Risk Factors",

author = "{Van Geloven}, Nan and Daniele Giardiello and Bonneville, {Edouard F.} and Lucy Teece and Ramspek, {Chava L.} and {Van Smeden}, Maarten and Snell, {Kym I.E.} and {Van Calster}, Ben and Maja Pohar-Perme and Riley, {Richard D.} and Hein Putter and Ewout Steyerberg",

note = "Publisher Copyright: {\textcopyright} Published by the BMJ Publishing Group Limited.",

year = "2022",

month = may,

day = "24",

doi = "10.1136/bmj-2021-069249",

language = "English",

volume = "377",

journal = "The BMJ",

issn = "0959-8146",

publisher = "BMJ Publishing Group",

}

TY - JOUR

T1 - Validation of prediction models in the presence of competing risks

T2 - a guide through modern methods

AU - Van Geloven, Nan

AU - Giardiello, Daniele

AU - Bonneville, Edouard F.

AU - Teece, Lucy

AU - Ramspek, Chava L.

AU - Van Smeden, Maarten

AU - Snell, Kym I.E.

AU - Van Calster, Ben

AU - Pohar-Perme, Maja

AU - Riley, Richard D.

AU - Putter, Hein

AU - Steyerberg, Ewout

N1 - Publisher Copyright: © Published by the BMJ Publishing Group Limited.

PY - 2022/5/24

Y1 - 2022/5/24

N2 - Thorough validation is pivotal for any prediction model before it can be advocated for use in medical practice. For time-to-event outcomes such as breast cancer recurrence, death from other causes is a competing risk. Model performance measures must account for such competing events. In this article, we present a comprehensive yet accessible overview of performance measures for this competing event setting, including the calculation and interpretation of statistical measures for calibration, discrimination, overall prediction error, and clinical usefulness by decision curve analysis. All methods are illustrated for patients with breast cancer, with publicly available data and R code.

AB - Thorough validation is pivotal for any prediction model before it can be advocated for use in medical practice. For time-to-event outcomes such as breast cancer recurrence, death from other causes is a competing risk. Model performance measures must account for such competing events. In this article, we present a comprehensive yet accessible overview of performance measures for this competing event setting, including the calculation and interpretation of statistical measures for calibration, discrimination, overall prediction error, and clinical usefulness by decision curve analysis. All methods are illustrated for patients with breast cancer, with publicly available data and R code.

KW - Humans

KW - Models, Statistical

KW - Proportional Hazards Models

KW - Risk Assessment

KW - Risk Factors

UR - http://www.scopus.com/inward/record.url?scp=85131105973&partnerID=8YFLogxK

U2 - 10.1136/bmj-2021-069249

DO - 10.1136/bmj-2021-069249

M3 - Article

C2 - 35609902

AN - SCOPUS:85131105973

SN - 0959-8146

VL - 377

JO - The BMJ

JF - The BMJ

M1 - e069249

ER -

Validation of prediction models in the presence of competing risks: a guide through modern methods

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this