A comparison of the beta-geometric model with landmarking for dynamic prediction of time to pregnancy

Rik van Eekelen*, Hein Putter, David J. McLernon, Marinus J. Eijkemans, Nan van Geloven

*Corresponding author for this work

Research output: Contribution to journalArticle

Abstract

We conducted a simulation study to compare two methods that have been recently used in clinical literature for the dynamic prediction of time to pregnancy. The first is landmarking, a semi-parametric method where predictions are updated as time progresses using the patient subset still at risk at that time point. The second is the beta-geometric model that updates predictions over time from a parametric model estimated on all data and is specific to applications with a discrete time to event outcome. The betageometric model introduces unobserved heterogeneity by modelling the chance of an event per discrete time unit according to a beta distribution. Due to selection of patients with lower chances as time progresses, the predicted probability of an event decreases over time. Both methods were recently used to develop models predicting the chance to conceive naturally. The advantages, disadvantages and accuracy of these two methods are unknown. We simulated time-to-pregnancy data according to different scenarios. We then compared the two methods by the following out-of-sample metrics: bias and root mean squared error in the average prediction, root mean squared error in individual predictions, Brier score and c statistic. We consider different scenarios including data-generating mechanisms for which the models are misspecified. We applied the two methods on a clinical dataset comprising 4999 couples. Finally, we discuss the pros and cons of the two methods based on our results and present
recommendations for use of either of the methods in different settings and (effective) sample sizes.
Original languageEnglish
JournalBiometrical Journal
Early online date18 Nov 2019
DOIs
Publication statusE-pub ahead of print - 18 Nov 2019

Fingerprint

Geometric Model
Pregnancy
Prediction
Mean Squared Error
Discrete-time
Effective Sample Size
Roots
Unobserved Heterogeneity
Semiparametric Methods
Scenarios
Beta distribution
Parametric Model
Statistic
Update
Simulation Study
Model
Metric
Unknown
Decrease
Unit

Keywords

  • beta-geometric model
  • Cox model
  • dynamic prediction
  • frailty
  • heterogeneity
  • landmarking
  • time to pregnancy

Cite this

A comparison of the beta-geometric model with landmarking for dynamic prediction of time to pregnancy. / van Eekelen, Rik; Putter, Hein; McLernon, David J.; Eijkemans, Marinus J.; van Geloven, Nan.

In: Biometrical Journal, 18.11.2019.

Research output: Contribution to journalArticle

@article{3601aa54b7cf4d16869f4e561f03c48e,
title = "A comparison of the beta-geometric model with landmarking for dynamic prediction of time to pregnancy",
abstract = "We conducted a simulation study to compare two methods that have been recently used in clinical literature for the dynamic prediction of time to pregnancy. The first is landmarking, a semi-parametric method where predictions are updated as time progresses using the patient subset still at risk at that time point. The second is the beta-geometric model that updates predictions over time from a parametric model estimated on all data and is specific to applications with a discrete time to event outcome. The betageometric model introduces unobserved heterogeneity by modelling the chance of an event per discrete time unit according to a beta distribution. Due to selection of patients with lower chances as time progresses, the predicted probability of an event decreases over time. Both methods were recently used to develop models predicting the chance to conceive naturally. The advantages, disadvantages and accuracy of these two methods are unknown. We simulated time-to-pregnancy data according to different scenarios. We then compared the two methods by the following out-of-sample metrics: bias and root mean squared error in the average prediction, root mean squared error in individual predictions, Brier score and c statistic. We consider different scenarios including data-generating mechanisms for which the models are misspecified. We applied the two methods on a clinical dataset comprising 4999 couples. Finally, we discuss the pros and cons of the two methods based on our results and presentrecommendations for use of either of the methods in different settings and (effective) sample sizes.",
keywords = "beta-geometric model, Cox model, dynamic prediction, frailty, heterogeneity, landmarking, time to pregnancy",
author = "{van Eekelen}, Rik and Hein Putter and McLernon, {David J.} and Eijkemans, {Marinus J.} and {van Geloven}, Nan",
year = "2019",
month = "11",
day = "18",
doi = "10.1002/bimj.201900155",
language = "English",
journal = "Biometrical Journal",
issn = "0323-3847",
publisher = "Wiley-VCH Verlag",

}

TY - JOUR

T1 - A comparison of the beta-geometric model with landmarking for dynamic prediction of time to pregnancy

AU - van Eekelen, Rik

AU - Putter, Hein

AU - McLernon, David J.

AU - Eijkemans, Marinus J.

AU - van Geloven, Nan

PY - 2019/11/18

Y1 - 2019/11/18

N2 - We conducted a simulation study to compare two methods that have been recently used in clinical literature for the dynamic prediction of time to pregnancy. The first is landmarking, a semi-parametric method where predictions are updated as time progresses using the patient subset still at risk at that time point. The second is the beta-geometric model that updates predictions over time from a parametric model estimated on all data and is specific to applications with a discrete time to event outcome. The betageometric model introduces unobserved heterogeneity by modelling the chance of an event per discrete time unit according to a beta distribution. Due to selection of patients with lower chances as time progresses, the predicted probability of an event decreases over time. Both methods were recently used to develop models predicting the chance to conceive naturally. The advantages, disadvantages and accuracy of these two methods are unknown. We simulated time-to-pregnancy data according to different scenarios. We then compared the two methods by the following out-of-sample metrics: bias and root mean squared error in the average prediction, root mean squared error in individual predictions, Brier score and c statistic. We consider different scenarios including data-generating mechanisms for which the models are misspecified. We applied the two methods on a clinical dataset comprising 4999 couples. Finally, we discuss the pros and cons of the two methods based on our results and presentrecommendations for use of either of the methods in different settings and (effective) sample sizes.

AB - We conducted a simulation study to compare two methods that have been recently used in clinical literature for the dynamic prediction of time to pregnancy. The first is landmarking, a semi-parametric method where predictions are updated as time progresses using the patient subset still at risk at that time point. The second is the beta-geometric model that updates predictions over time from a parametric model estimated on all data and is specific to applications with a discrete time to event outcome. The betageometric model introduces unobserved heterogeneity by modelling the chance of an event per discrete time unit according to a beta distribution. Due to selection of patients with lower chances as time progresses, the predicted probability of an event decreases over time. Both methods were recently used to develop models predicting the chance to conceive naturally. The advantages, disadvantages and accuracy of these two methods are unknown. We simulated time-to-pregnancy data according to different scenarios. We then compared the two methods by the following out-of-sample metrics: bias and root mean squared error in the average prediction, root mean squared error in individual predictions, Brier score and c statistic. We consider different scenarios including data-generating mechanisms for which the models are misspecified. We applied the two methods on a clinical dataset comprising 4999 couples. Finally, we discuss the pros and cons of the two methods based on our results and presentrecommendations for use of either of the methods in different settings and (effective) sample sizes.

KW - beta-geometric model

KW - Cox model

KW - dynamic prediction

KW - frailty

KW - heterogeneity

KW - landmarking

KW - time to pregnancy

U2 - 10.1002/bimj.201900155

DO - 10.1002/bimj.201900155

M3 - Article

JO - Biometrical Journal

JF - Biometrical Journal

SN - 0323-3847

ER -