Pith. sign in

REVIEW 3 major objections 6 minor 300 references

This paper claims that discretizing follow-up time turns joint prediction of a terminal event and a longitudinal marker into a single Bayesian model whose marker trajectory is conditional on being alive—and demonstrates it on quality of lif

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-02 07:27 UTC pith:O2QPS2Y5

load-bearing objection A well-executed framework paper with a real clinical motivation, but the partial-information missingness strategy has an unaddressed gap that directly affects the SCD-HeFT death-hazard terms. the 3 major comments →

arxiv 2607.09931 v2 pith:O2QPS2Y5 submitted 2026-07-10 stat.ME

Toward Joint Prediction of a Longitudinal Marker and a Terminal Event: A bivariate discrete-time framework

Stephanie Armbruster , Daniel Kramer , Rui Duan , Rajarshi Mukherjee , Sebastien Haneuse , Harrison Reeder This is my paper
classification stat.ME MSC 62N0162F1562P10
keywords Discrete-time partitionTerminal event outcomeLongitudinal marker trajectoryRandom effectAutoregressionShared frailtyBayesian estimationHamiltonian Monte Carlo
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that partitioning follow-up time into discrete intervals makes it possible to model death and a repeatedly measured health marker as one bivariate process, and to issue patient-specific joint predictions of both. The marker trajectory is defined conditional on being alive at each interval—a 'partly conditional' mean—so predictions do not imagine an immortal cohort in which death never happens. The framework lets the death hazard depend on recent marker values, with either constant or time-varying strength, plus a shared patient frailty, and estimates all of this in a Bayesian way. Applied to heart-failure patients with implantable defibrillators, it produces distinct predicted journeys: good quality of life with low mortality, stable poor quality of life with rising mortality, rapid deterioration, and high baseline burden. If this works, a clinician could weigh mortality against quality of life in a single conversation instead of relying on separate univariate predictions.

Core claim

The central claim is that the bivariate discrete-time framework provides joint prediction of time to terminal event and longitudinal marker trajectory by conditioning each interval's marker distribution on survival through that interval, thereby avoiding the implicit 'immortal cohort' extrapolation of standard joint models. The paper shows in simulations that correctly specified models recover regression parameters with near-nominal coverage and that marginal Bayesian model-selection criteria can pick a parsimonious model from a candidate set. In the SCD-HeFT application, the base model with global dependence and shared frailty is preferred, and patient-specific posterior predictions span a

What carries the argument

The key object is the bivariate discrete-time process (N_{k,i}, Y_{k,i}) built on a partition of continuous study time. At each interval k, the joint density factorizes as a Bernoulli terminal-event indicator with hazard pi_{k,i} = P(N_{k,i}=1 | N_{k-1,i}=0, history) times a marker distribution f_Y(y_{k,i} | N_{k,i}=0, history), whose mean mu_{k,i} is the partly conditional trajectory. This sequential conditioning on survival is what makes the marker trajectory interpretable as the expected value among patients still alive, and it lets the terminal hazard depend on lagged marker values through global and local dependence terms, on a shared frailty, and on autoregressive and random-effect str

Load-bearing premise

The whole analysis hinges on the assumption that a patient who skips a follow-up visit is, on average, no sicker than one who attends once baseline information is accounted for—if missed visits actually reflect declining health, the predicted trajectories and the link between quality of life and mortality will be biased.

What would settle it

In the SCD-HeFT data, fit a model for whether a patient missed a scheduled visit as a function of baseline covariates, the last observed marker value, and subsequent death status; if missingness predicts death beyond the observed marker history, non-informative missingness is false and the framework's estimates are suspect. Alternatively, a sensitivity analysis that systematically imputes missed markers as worse than the last observed value should materially change the joint predictions if the assumption is load-bearing.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Joint posterior draws respect death: each predicted marker trajectory stops at the predicted time of death, so summaries such as 'expected quality of life at 24 months' are averages over patients alive at 24 months.
  • For ICD patients, the framework separates those who will maintain good quality of life with low mortality from those who will deteriorate or die, information that can directly inform ICD continuation, deactivation, or palliative care discussions.
  • Marginal BIC, WAIC, and ELPD-based model selection favor parsimonious specifications in simulations, giving analysts a principled way to choose among global dependence, local dependence, autoregression, and shared frailty options.
  • The partial-information strategy for intermittent missingness avoids the artificial autocorrelation introduced by last-observation-carried-forward and linear interpolation in the autoregressive parameter.
  • Correctly specified and flexibly parameterized models show small bias and near-nominal credible-interval coverage in the simulation study, while misspecification of the terminal event submodel or dependence structure biases both submodels.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • My inference: the same survival-conditional factorization should extend naturally to competing risks or multiple longitudinal markers, because the argument does not depend on the number of markers or event types; the paper mentions such extensions as future work.
  • My inference: the choice of discrete-time partition is not just a technical nuisance but a substantive decision about clinical timescale, so sensitivity analyses over interval widths could matter as much as model selection.
  • My inference: because the longitudinal predictions are partly conditional, a reader must interpret them as 'if you are alive then,' not as a unconditional forecast; a decision aid built on these predictions should state that interpretation explicitly.
  • My inference: the paper's own observation that no joint predictive-accuracy metric exists points to a concrete testable next step—define a proper scoring rule for pairs of survival time and marker trajectory and compare the bivariate model against univariate benchmarks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes a Bayesian bivariate discrete-time framework for jointly modeling a longitudinal marker and a terminal event, with the explicit goal of patient-specific joint prediction. The key design feature is discretizing time so that the longitudinal mean is defined conditionally on being alive, i.e., a partly-conditional trajectory, thereby avoiding implicit extrapolation of the marker beyond death. The framework allows flexible dependence structures: global time-invariant and local time-varying dependence of the hazard on lagged marker values, autoregressive marker dynamics, and a shared frailty. Estimation is carried out in Stan, with model selection via marginal BIC, WAIC, and PSIS leave-one-out criteria. A simulation study shows good parameter recovery and model-selection behavior for correctly specified or slightly overparameterized models. The SCD-HeFT application produces joint posterior predictions of MLHFQ quality-of-life trajectories and mortality for hypothetical ICD patients, and the authors discuss how these predictions could inform ICD continuation/deactivation decisions.

Significance. If the identified gaps are resolved, the framework would be a useful addition to the joint-modeling and dynamic-prediction literature, particularly for clinical settings where death truncates the longitudinal trajectory and both outcomes are of intrinsic interest. The partly-conditional interpretation of the longitudinal mean is clinically natural, and the unified Bayesian computation in Stan with an accompanying R package is a practical strength. The paper is also honest about its scope, presenting the work as a framework rather than a fully validated prediction tool. However, the central predictive claims currently rest on an incompletely specified likelihood when lagged markers are missing, and on a strong untested missingness assumption in the application. These issues are load-bearing for the SCD-HeFT results and for the joint predictions in Figures 5–6.

major comments (3)
  1. [§3.5, Eq. (1); Appendix I.1] The partial-information likelihood is not computable when the lagged marker is missing in the death interval. Eq. (1) and the Appendix I.1 product include the death contribution π_{k_r,i}^{Δ_i} unconditionally, even if M_{k_r-1,i}=0. Since π_{k,i} contains global/local dependence on Y_{k-1,i}, the death term then depends on an unobserved value. The paper never states whether such deaths are omitted, whether a last-observation value is substituted, or whether integration over missing Y is performed. With 87 deaths in the partial-information analytic sample, this is not a corner case; it directly affects the estimated dependence ϑ and the joint predictions used in Figures 5–6. The authors must specify the evaluable likelihood contribution and, ideally, conduct a sensitivity analysis.
  2. [§8.2, Appendix I.1] The analysis assumes that intermittent missingness of the MLHFQ score is non-informative: independent of time to death and the marker value conditional on baseline covariates. The paper reports that 62.3% of patients missed at least one MLHFQ question. If sicker patients are more likely to miss visits, the estimated partly-conditional trajectories, the dependence parameter ϑ, and the joint predictions for ICD decision-making may be biased. The assumption is stated but not tested, and no sensitivity analysis is provided for informative missingness. A pattern-mixture or shared-parameter extension would be one way to assess robustness; at minimum, a discussion of the direction and plausible magnitude of bias is needed.
  3. [§7, Tables A13–A14] Although the stated primary goal is prediction, the simulation study evaluates only parameter recovery and model selection among candidate models from the same bivariate discrete-time family. There is no assessment of predictive calibration or discrimination (e.g., Brier score, AUC, interval scores for the joint outcome) and no comparison with existing dynamic-prediction alternatives such as landmarking, joint longitudinal–survival models, or separate univariate predictions. The claim that the framework 'provides' joint predictions that are clinically useful would be materially strengthened by evidence that the posterior predictive distributions are calibrated and at least competitive with simpler approaches. This is a substantive gap, not merely a presentation issue.
minor comments (6)
  1. [§2.2] Typographical errors: 'whcih' and 'trajectoryies' in Section 2.3. 'SCD-Heft' also appears in Section 6.1; use consistent SCD-HeFT.
  2. [Figure A3] The flow diagram reports N=784 for Case 3 (linear imputation), exceeding the complete-case base N=612. This appears to be an error or a different sample construction; please explain or correct.
  3. [§4.1] The prior for σ_α is described as 'weakly informative,' but the actual value is not reported in the main text or simulation summary. The simulations use σ_α=20 in Table A5; the application likely uses a different value. Report the values used for the SCD-HeFT analysis.
  4. [Table 2] The notation RCS(k=3) is not defined in the table footnote. Clarify that k denotes the number of knots and state the knot locations or that they are placed at quantiles.
  5. [Figures 5–6] The figures show 50 predicted trajectories with 'a random subset in bold,' but the legend does not identify which subset. A simple statement in the caption or legend would improve readability.
  6. [§3.4.3] The shared frailty linear predictor αγ_i is added on the link scale in the hazard but on the mean scale in the longitudinal submodel. This asymmetry is not discussed. A short explanation of why this is the intended parameterization would help readers.

Circularity Check

0 steps flagged

No significant circularity; the partly-conditional property is built into the model definition and the Nevo et al. citation is not load-bearing.

full rationale

The paper's central claim—that discretizing time yields partly-conditional longitudinal predictions without extrapolating past death—is a direct consequence of the model construction, not a derived empirical result. In Section 3.2, Y_{k,i} is defined only when N_{k,i}=0, and in Section 3.3, the estimand µ_{k,i} is explicitly defined as E(Y_{k,i}|N_{k,i}=0,H_{k,i}); the 'partly-conditional' interpretation is therefore the target of estimation, not a prediction smuggled in from external inputs. The self-cited Nevo et al. (2022) is used for the global/local dependence formulation and for observation-attribution options, but the paper's estimation strategy, simulation study, and SCD-HeFT application are implemented and evaluated independently here, so the self-citation is not load-bearing. The only substantive concern is in Appendix I.1, where the partial-information likelihood retains π^{Δ_i}_{k_i^r,i} even if the lagged marker Y_{k_i^r-1,i} is missing, making that term not directly computable from observed data; however, this is an estimation gap or missing-data assumption issue, not circularity, because nothing in the framework's derivation reduces its predictions to its fitted inputs. Overall, the derivation chain is self-contained and no circular step is exhibited.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 1 invented entities

The framework is a modeling construct, so the ledger lists the distributional and missingness assumptions required for its validity plus user-chosen design parameters (partition, attribution, priors) that affect the estimands. No new physical entities are introduced; the shared frailty is a latent statistical variable without independent evidence.

free parameters (3)
  • Discrete-time partition (6-month intervals for SCD-HeFT) = 6 months (0, 6, ..., 72 months)
    Chosen to match the SCD-HeFT visit schedule; determines the timescale of hazards and lag distances. The paper gives guidance but does not fit the partition to the data.
  • Attribution rule for continuous-time observations = Ceiling (carry time-to-death forward)
    Chosen for the SCD-HeFT analysis to retain all observations; affects interval assignment of marker values and death times, and can alter hazard and dependence estimates.
  • Prior standard deviation for scale multiplier α (σ_α) = Not explicitly reported; prior α ∼ N(0, σ_α²)
    A weakly informative prior on the shared-frailty weight; its scale is a user-chosen hyperparameter that influences shrinkage of α and is part of the Bayesian model specification.
axioms (5)
  • domain assumption Non-informative right censoring: (N_{k,i}, Y_{k,i}) ⊥ C_i | X_i
    Used to factor the observed-data likelihood (Appendix D); standard in survival analysis but not directly testable from data.
  • domain assumption Non-informative intermittent missingness: (N_{k,i}, Y_{k,i}) ⊥ M_{k,i} | X_i
    Assumed for the SCD-HeFT application (Section 8.2, Appendix I.1); required for validity of the partial-information likelihood when visits are missed.
  • domain assumption Sequential factorization of the joint density: p(N_k,Y_k|N_{k-1}=0,H_k) = π_k if N_k=1 and (1-π_k)f_Y(Y_k|N_k=0,H_k) if N_k=0
    Defines the joint model (Section 3.3); implies no contemporaneous residual association between Y_k and N_k beyond the hazard's dependence on history.
  • domain assumption The longitudinal marker follows a parametric distribution whose mean is the partly-conditional mean (e.g., Gaussian with identity link in the implementation)
    Required for the likelihood; the paper illustrates several distributions (Appendix B), but the simulation and application use Gaussian/identity link.
  • ad hoc to paper Identifiability of latent factors (b_i, γ_i) via subtraction of Z_i^T b_i and γ_i from lagged marker terms in dependence structures
    The authors introduce this parameterization to ensure identifiability and to preserve a marginal interpretation of latent factors (Section 3.4.4). It is a modeling constraint specific to this framework.
invented entities (1)
  • Shared frailty γ_i no independent evidence
    purpose: Captures unobserved patient-specific residual dependence between the longitudinal marker trajectory and the terminal event hazard beyond observed covariates and direct marker dependence.
    A latent variable with no directly falsifiable measurement; its existence is inferred only through model fit. It is a standard construct in joint modeling, not a newly discovered physical entity, but it is a postulated statistical entity without independent external evidence.

pith-pipeline@v1.3.0-alltime-deepseek · 36152 in / 18244 out tokens · 196475 ms · 2026-08-02T07:27:03.259704+00:00 · methodology

0 comments
read the original abstract

Sudden cardiac death (SCD) is a leading cause of death in the U.S. Patients at elevated risk of SCD are primarily treated with an implantable cardioverter-defibrillator (ICD), which may prevent death from cardiovascular causes but may cause severe side effects, such as reduced quality of life from shock-induced pain. Decisions about ICD treatment therefore involve complex personal trade-offs across multiple health events, including mortality and quality of life. While prediction tools could help weigh these trade-offs, they commonly focus on univariate outcomes; at best, they treat other clinical endpoints as inputs, so trade-offs cannot be directly informed. To address this, we propose a novel general Bayesian framework that jointly models a terminal event and a longitudinal marker as a bivariate process over discrete time, for settings where prediction is the primary goal. Discretization of study time lets the framework capture the dynamic interplay between outcomes while avoiding implicit extrapolation beyond truncation by a terminal event. The framework flexibly accommodates phenomena arising in applied contexts, including global time-invariant and local time-dependent dependence structures between the terminal event and the longitudinal marker, and latent association via a shared frailty term. Estimation proceeds via the Bayesian paradigm, yielding patient-specific joint posterior predictions for the time to terminal event and the future marker trajectory. We introduce the framework with a focus on its modeling flexibility, provide guidance on discretization and on Bayesian model construction and selection, and discuss insights from the joint posterior predictions. Finally, we demonstrate the framework's clinical relevance and practicality for SCD and ICD therapy using data from the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT), an important ICD-related benchmark trial.

Figures

Figures reproduced from arXiv: 2607.09931 by Daniel Kramer, Harrison Reeder, Rajarshi Mukherjee, Rui Duan, Sebastien Haneuse, Stephanie Armbruster.

Figure 1
Figure 1. Figure 1: Stratified Kaplan-Meier estimate for survival probability for all-cause mortality and the [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of mean MLHFQ score trajectories for different patient cohorts and of MLHFQ [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 2
Figure 2. Figure 2: Bivariate discrete-time process for terminal event indicator [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Bivariate discrete-time process for terminal event indicator [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Graphical representation of the bivariate discrete-time framework; [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of the different approaches (Nearest Neighbor (NN), Ceiling, and Floor) to [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of the different approaches (Nearest Neighbor (NN), Ceiling, and Floor) to [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 5
Figure 5. Figure 5: Patient-specific conditional joint posterior predictions for the base model under partial [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Patient-specific characterization of conditional joint posterior predictive distribution for [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Patient-specific conditional joint posterior predictions for the base model under partial [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Conditional joint prediction for patient 2 under partial information strategy for intermittent [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

300 extracted references · 5 linked inside Pith

  1. [1]

    2013 , publisher=

    Counting processes and Survival Analysis , author=. 2013 , publisher=

    2013

  2. [2]

    International journal of forecasting , volume=

    Another look at measures of forecast accuracy , author=. International journal of forecasting , volume=. 2006 , publisher=

    2006

  3. [3]

    Statistical Methods in Medical Research , volume=

    Multi-state models for event history analysis , author=. Statistical Methods in Medical Research , volume=. 2002 , publisher=

    2002

  4. [4]

    Statistics in Medicine , volume=

    Regression models for expected length of stay , author=. Statistics in Medicine , volume=. 2016 , publisher=

    2016

  5. [5]

    Lifetime data analysis , volume=

    Regression analysis of restricted mean survival time based on pseudo-observations , author=. Lifetime data analysis , volume=. 2004 , publisher=

    2004

  6. [6]

    Biometrics , volume=

    Accelerated failure time models for semi-competing risks data in the presence of complex censoring , author=. Biometrics , volume=. 2017 , publisher=

    2017

  7. [7]

    Journal of the Royal Statistical Society Series C: Applied Statistics , volume=

    Bayesian semiparametric analysis of semicompeting risks data: investigating hospital readmission after a pancreatic cancer diagnosis , author=. Journal of the Royal Statistical Society Series C: Applied Statistics , volume=. 2015 , publisher=

    2015

  8. [8]

    Structural Equation Modeling: A Multidisciplinary Journal , volume=

    The GRoLTS-checklist: guidelines for reporting on latent trajectory studies , author=. Structural Equation Modeling: A Multidisciplinary Journal , volume=. 2017 , publisher=

    2017

  9. [9]

    Statistics in Medicine , volume=

    Joint modeling of repeated multivariate cognitive measures and competing risks of dementia and death: a latent process and latent class approach , author=. Statistics in Medicine , volume=. 2016 , publisher=

    2016

  10. [10]

    Biometrics , volume=

    Joint latent class model for longitudinal data and interval-censored semi-competing events: Application to dementia , author=. Biometrics , volume=. 2016 , publisher=

    2016

  11. [11]

    2012 , publisher=

    Joint models for longitudinal and time-to-event data: With applications in R , author=. 2012 , publisher=

    2012

  12. [12]

    and Davidian, Marie , journal =

    Tsiatis, Anastasios A. and Davidian, Marie , journal =. JOINT MODELING OF LONGITUDINAL AND TIME-TO-EVENT DATA: AN OVERVIEW , urldate =

  13. [13]

    Statistical Methods in Medical Research , volume=

    Joint latent class models for longitudinal and time-to-event data: a review , author=. Statistical Methods in Medical Research , volume=. 2014 , publisher=

    2014

  14. [14]

    and Tsiatis, Anastasios A

    Wulfsohn, Michael S. and Tsiatis, Anastasios A. , journal =. A Joint Model for Survival and Longitudinal Data Measured with Error , urldate =

  15. [15]

    Journal of the Royal Statistical Society Series A: Statistics in Society , volume=

    Joint modelling of longitudinal outcome and interval-censored competing risk dropout in a schizophrenia clinical trial , author=. Journal of the Royal Statistical Society Series A: Statistics in Society , volume=. 2012 , publisher=

    2012

  16. [16]

    Journal of the American Statistical Association , volume=

    Latent class models for joint analysis of longitudinal biomarker and event process data: application to longitudinal prostate-specific antigen readings and prostate cancer , author=. Journal of the American Statistical Association , volume=. 2002 , publisher=

    2002

  17. [17]

    Statistics in Medicine , volume=

    Landmarking 2.0: Bridging the gap between joint models and landmarking , author=. Statistics in Medicine , volume=. 2022 , publisher=

    2022

  18. [18]

    Biometrical Journal , volume=

    Comparison of joint modeling and landmarking for dynamic prediction under an illness-death model , author=. Biometrical Journal , volume=. 2017 , publisher=

    2017

  19. [19]

    Applications to survival and CD4 counts in patients with AIDS , author=

    Modeling the relationship of survival to longitudinal data measured with error. Applications to survival and CD4 counts in patients with AIDS , author=. Journal of the American Statistical Association , volume=. 1995 , publisher=

    1995

  20. [20]

    Briefings in Bioinformatics , volume=

    A review on longitudinal data analysis with random forest , author=. Briefings in Bioinformatics , volume=. 2023 , publisher=

    2023

  21. [21]

    arXiv preprint arXiv:1706.05098 , year=

    An overview of multi-task learning in deep neural networks , author=. arXiv preprint arXiv:1706.05098 , year=

    Pith/arXiv arXiv

  22. [22]

    Joint European Conference on Machine Learning and Knowledge Discovery in Databases , pages=

    A general machine learning framework for survival analysis , author=. Joint European Conference on Machine Learning and Knowledge Discovery in Databases , pages=. 2020 , organization=

    2020

  23. [23]

    ACM Computing Surveys (CSUR) , volume=

    Machine learning for survival analysis: A survey , author=. ACM Computing Surveys (CSUR) , volume=. 2019 , publisher=

    2019

  24. [24]

    Proceedings of the 1st Machine Learning for Healthcare Conference , pages =

    Deep Survival Analysis , author =. Proceedings of the 1st Machine Learning for Healthcare Conference , pages =. 2016 , editor =

    2016

  25. [25]

    IEEE Transactions on Biomedical Engineering , volume=

    Dynamic-deephit: A deep learning approach for dynamic survival analysis with competing risks based on longitudinal data , author=. IEEE Transactions on Biomedical Engineering , volume=. 2020 , publisher=

    2020

  26. [26]

    Statistics in Medicine , volume=

    Model-assisted analyses of longitudinal, ordinal outcomes with absorbing states , author=. Statistics in Medicine , volume=. 2022 , publisher=

    2022

  27. [27]

    Scandinavian Journal of Statistics , volume=

    A semi-parametric transformation frailty model for semi-competing risks survival data , author=. Scandinavian Journal of Statistics , volume=. 2017 , publisher=

    2017

  28. [28]

    Circulation: Cardiovascular Quality and Outcomes , volume=

    Joint shock/death risk prediction model for patients considering implantable cardioverter-defibrillators: a secondary analysis of the SCD-HeFT trial , author=. Circulation: Cardiovascular Quality and Outcomes , volume=. 2019 , publisher=

    2019

  29. [29]

    Biometrics , volume=

    Penalized estimation of frailty-based illness--death models for semi-competing risks , author=. Biometrics , volume=. 2023 , publisher=

    2023

  30. [30]

    Biostatistics , volume=

    Causal inference for semi-competing risks data , author=. Biostatistics , volume=. 2022 , publisher=

    2022

  31. [31]

    Journal of the American Statistical Association , volume=

    Marginalized frailty-based illness-death model: application to the UK-Biobank survival data , author=. Journal of the American Statistical Association , volume=. 2021 , publisher=

    2021

  32. [32]

    Biometrics , volume=

    Modeling semi-competing risks data as a longitudinal bivariate process , author=. Biometrics , volume=. 2022 , publisher=

    2022

  33. [33]

    Journal of the American Statistical Association , volume=

    Regression models and multivariate life tables , author=. Journal of the American Statistical Association , volume=. 2021 , publisher=

    2021

  34. [34]

    Biometrics , volume=

    Statistical analysis of illness--death processes and semicompeting risks data , author=. Biometrics , volume=. 2010 , publisher=

    2010

  35. [35]

    Statistical Science , volume=

    Longitudinal data with follow-up truncated by death: match the analysis method to research aims , author=. Statistical Science , volume=. 2009 , publisher=

    2009

  36. [36]

    Journal of Clinical Oncology , volume=

    Analysis of survival by tumor response , author=. Journal of Clinical Oncology , volume=

  37. [37]

    Lifetime data analysis , volume=

    Dynamic predicting by landmarking as an alternative for multi-state modeling: an application to acute lymphoid leukemia data , author=. Lifetime data analysis , volume=. 2008 , publisher=

    2008

  38. [38]

    Statistics in Medicine , volume=

    Dynamic prediction by landmarking in competing risks , author=. Statistics in Medicine , volume=. 2013 , publisher=

    2013

  39. [39]

    Journal of the American Statistical Association , volume=

    A proportional hazards model for the subdistribution of a competing risk , author=. Journal of the American Statistical Association , volume=. 1999 , publisher=

    1999

  40. [40]

    Biometrika , volume=

    On semi-competing risks data , author=. Biometrika , volume=. 2001 , publisher=

    2001

  41. [41]

    Lifetime data analysis , volume=

    Semicompeting risks in aging research: methods, issues and needs , author=. Lifetime data analysis , volume=. 2014 , publisher=

    2014

  42. [42]

    Statistical Methods in Medical Research , volume=

    Time-to-event analysis when the event is defined on a finite time interval , author=. Statistical Methods in Medical Research , volume=. 2020 , publisher=

    2020

  43. [43]

    Biometrics , pages=

    A penalized likelihood approach for arbitrarily censored and truncated data: application to age-specific incidence of dementia , author=. Biometrics , pages=. 1998 , publisher=

    1998

  44. [44]

    JAMA Network Open , volume=

    Development and assessment of a model for predicting individualized outcomes in patients with oropharyngeal cancer , author=. JAMA Network Open , volume=. 2021 , publisher=

    2021

  45. [45]

    Statistics in Medicine , volume=

    Multi-state models for colon cancer recurrence and death with a cured fraction , author=. Statistics in Medicine , volume=. 2014 , publisher=

    2014

  46. [46]

    Statistics in Medicine , volume=

    Matching with time-dependent treatments: A review and look forward , author=. Statistics in Medicine , volume=. 2020 , publisher=

    2020

  47. [47]

    Biostatistics , volume=

    EM algorithms for fitting multistate cure models , author=. Biostatistics , volume=. 2019 , publisher=

    2019

  48. [48]

    Scientific reports , volume=

    Joint modelling of colorectal cancer recurrence and death after resection using multi-state model with cured fraction , author=. Scientific reports , volume=. 2021 , publisher=

    2021

  49. [49]

    Computer Methods and Programs in Biomedicine , volume=

    Artificial intelligence based personalized predictive survival among colorectal cancer patients , author=. Computer Methods and Programs in Biomedicine , volume=. 2023 , publisher=

    2023

  50. [50]

    Computational Statistics and Applications , year=

    Dependent dirichlet processes for analysis of a generalized shared frailty model , author=. Computational Statistics and Applications , year=

  51. [51]

    A Bayesian semiparametric temporally-stratified proportional hazards model with spatial frailties , author=

  52. [52]

    Cancer Informatics , volume=

    Comparing Individualized Survival Predictions From Random Survival Forests and Multistate Models in the Presence of Missing Data: A Case Study of Patients With Oropharyngeal Cancer , author=. Cancer Informatics , volume=. 2023 , publisher=

    2023

  53. [53]

    Scandinavian Journal of Statistics , volume=

    Dynamic prediction by landmarking in event history analysis , author=. Scandinavian Journal of Statistics , volume=. 2007 , publisher=

    2007

  54. [54]

    Journal of the American Statistical Association , volume=

    The robust inference for the Cox proportional hazards model , author=. Journal of the American Statistical Association , volume=. 1989 , publisher=

    1989

  55. [55]

    Statistics in Medicine , volume=

    Individual frailty excess hazard models in cancer epidemiology , author=. Statistics in Medicine , volume=. 2023 , publisher=

    2023

  56. [56]

    Journal of Statistical Planning and Inference , volume=

    On predictive causality in longitudinal studies , author=. Journal of Statistical Planning and Inference , volume=. 1993 , publisher=

    1993

  57. [57]

    Statistics in Medicine , volume=

    Plotting summary predictions in multistate survival models: probabilities of relapse and death in remission for bone marrow transplantation patients , author=. Statistics in Medicine , volume=. 1993 , publisher=

    1993

  58. [58]

    Statistics in Medicine , volume=

    Individualized dynamic prediction of survival with the presence of intermediate events , author=. Statistics in Medicine , volume=. 2019 , publisher=

    2019

  59. [59]

    Circulation: Heart Failure , volume=

    Multistate model to predict heart failure hospitalizations and all-cause mortality in outpatients with heart failure with reduced ejection fraction: Model derivation and external validation , author=. Circulation: Heart Failure , volume=. 2016 , publisher=

    2016

  60. [60]

    Biometrics , volume=

    Bayesian path specific frailty models for multi-state survival data with applications , author=. Biometrics , volume=. 2015 , publisher=

    2015

  61. [61]

    Biostatistics , volume=

    Joint modeling of recurrent events and survival: a Bayesian non-parametric approach , author=. Biostatistics , volume=. 2020 , publisher=

    2020

  62. [62]

    Statistics in Medicine , volume=

    Design and analysis of nested case--control studies for recurrent events subject to a terminal event , author=. Statistics in Medicine , volume=. 2019 , publisher=

    2019

  63. [63]

    Statistics in Medicine , volume=

    Analysis of longitudinal studies with death and drop-out: a case study , author=. Statistics in Medicine , volume=. 2004 , publisher=

    2004

  64. [64]

    2011 , publisher=

    Dynamic prediction in clinical survival analysis , author=. 2011 , publisher=

    2011

  65. [65]

    American Journal of Epidemiology , volume=

    Use of repeated blood pressure and cholesterol measurements to improve cardiovascular disease risk prediction: an individual-participant-data meta-analysis , author=. American Journal of Epidemiology , volume=. 2017 , publisher=

    2017

  66. [66]

    Statistics in Medicine , volume=

    The use of repeated blood pressure measures for cardiovascular risk prediction: a comparison of statistical models in the ARIC study , author=. Statistics in Medicine , volume=. 2017 , publisher=

    2017

  67. [67]

    American Journal of Epidemiology , volume=

    Landmark models for optimizing the use of repeated measurements of risk factors in electronic health records to predict future disease risk , author=. American Journal of Epidemiology , volume=. 2018 , publisher=

    2018

  68. [68]

    Statistics in Medicine , volume=

    Penalized regression calibration: A method for the prediction of survival outcomes using complex longitudinal and high-dimensional data , author=. Statistics in Medicine , volume=. 2021 , publisher=

    2021

  69. [69]

    Biometrics , volume=

    Dynamic predictions and prospective accuracy in joint models for longitudinal and time-to-event data , author=. Biometrics , volume=. 2011 , publisher=

    2011

  70. [70]

    Statistical Methods in Medical Research , volume=

    Random survival forests with multivariate longitudinal endogenous covariates , author=. Statistical Methods in Medical Research , volume=. 2023 , publisher=

    2023

  71. [71]

    BMC medical research methodology , volume=

    Random survival forests for dynamic predictions of a time-to-event outcome using a longitudinal biomarker , author=. BMC medical research methodology , volume=. 2021 , publisher=

    2021

  72. [72]

    Random survival forests , author=

  73. [73]

    Statistics in Medicine , volume=

    Deep learning for the dynamic prediction of multivariate longitudinal and survival data , author=. Statistics in Medicine , volume=. 2022 , publisher=

    2022

  74. [74]

    Biostatistics , volume=

    Directly parameterized regression conditioning on being alive: analysis of longitudinal data truncated by deaths , author=. Biostatistics , volume=. 2005 , publisher=

    2005

  75. [75]

    Statistics in Medicine , volume=

    Joint modeling quality of life and survival using a terminal decline model in palliative care studies , author=. Statistics in Medicine , volume=. 2013 , publisher=

    2013

  76. [76]

    BMC Medical Research Methodology , volume=

    When a joint model should be preferred over a linear mixed model for analysis of longitudinal health-related quality of life data in cancer clinical trials , author=. BMC Medical Research Methodology , volume=. 2023 , publisher=

    2023

  77. [77]

    Journal of Clinical Epidemiology , volume=

    Analytical results in longitudinal studies depended on target of inference and assumed mechanism of attrition , author=. Journal of Clinical Epidemiology , volume=. 2015 , publisher=

    2015

  78. [78]

    Biostatistics , volume=

    Methods for handling longitudinal outcome processes truncated by dropout and death , author=. Biostatistics , volume=. 2018 , publisher=

    2018

  79. [79]

    Statistics in Medicine , volume=

    Weighted estimating equations for longitudinal studies with death and non-monotone missing time-dependent covariates and outcomes , author=. Statistics in Medicine , volume=. 2008 , publisher=

    2008

  80. [80]

    Statistics in Medicine , volume=

    Time-varying effect modeling with longitudinal data truncated by death: conditional models, interpretations, and inference , author=. Statistics in Medicine , volume=. 2016 , publisher=

    2016

Showing first 80 references.